Geir Kjetil Ferkingstad Sandve

• Ndagijimana, Similien; Semakula, Muhammed; Ishimwe, Christa; Sebakunzi, Thierry; Umuhire, Vivianne Niyonkuru & Hirwa, Sylvain Muzungu [Vis alle 24 forfattere av denne artikkelen] (2026). Spatial and spatio-temporal modelling of climate variability and under-five diarrhoeal disease in Rwanda. Environmental Research: Health. 4(3), s. 1–13. doi: 10.1088/2752-5309/ae8322. Fulltekst i vitenarkiv Vis sammendrag

  Background. Diarrhoeal disease remains a leading cause of morbidity and mortality among children under five years (U5) in Rwanda, contributing substantially to healthcare utilisation and imposing considerable economic burdens on households and the health system. National estimates indicate that approximately 14.3% of children under five experienced diarrhoea in the two weeks preceding the 2019–2020 Rwanda Demographic and Health Survey, and diarrhoea has been reported as the third leading cause of death in this age group. Climate variability may influence the risk of diarrhoeal disease, but evidence on how specific climatic factors affect diarrhoeal incidence across Rwanda’s diverse ecological zones and seasons remains limited. This study examined associations between maximum temperature, minimum temperature, rainfall, and relative humidity and U5 diarrhoeal incidence at the sector level in Rwanda. Methods. Monthly counts of U5 diarrhoeal cases reported by health facilities across 416 administrative sectors in Rwanda from January 2015 to December 2024 were analysed together with satellite-derived climate data. Spatio-temporal statistical models were used to evaluate associations between standardised climate variables and diarrhoeal incidence rates while accounting for geographic and seasonal variation. Model comparison relied on both goodness-of-fit metrics and out-of-sample predictive diagnostics. Results. Diarrhoeal incidence showed clear spatial and seasonal patterns, with persistently higher rates in northern and eastern Rwanda. In the final model maximum temperature was positively associated with increased diarrhoeal incidence, with a one-standard-deviation increase corresponding to a 5.6% rise in the estimated incidence rate ratio (RR = 1.056; 95% CrI: 1.02–1.09). Relative humidity showed a protective association (RR ≈ 0.919; 95% CrI: 0.89–0.94), while rainfall showed limited immediate effects. Conclusion. Under-five diarrhoeal incidence in Rwanda showed marked spatial and seasonal variability. Maximum temperature was positively associated with diarrhoeal incidence rates, while relative humidity showed an inverse association that may partly reflect unmeasured confounding by drought, water access, and WASH infrastructure. External validation is needed before the modelling framework can be applied operationally. These findings highlight the importance of preparedness during hotter periods and strengthened prevention and surveillance efforts in high-risk areas.

• Victoriano, Margarida; Pavlovic, Milena; Sandve, Geir Kjetil Ferkingstad; Oliveira, Hélder P.; Rocha, Artur & Greiff, Victor (2026). From virtual experiments to biomedical insight with synthetic data. Nature Machine Intelligence. 8, s. 866–879. doi: 10.1038/s42256-026-01244-6. Fulltekst i vitenarkiv
• Ferenc, Katalin Terezia; Martini, Lorenzo; Rauluseviciute, Ieva; Sandve, Geir Kjetil Ferkingstad & Mathelier, Anthony (2026). inMOTIFin: a lightweight end-to-end simulation software for regulatory sequences. Bioinformatics. ISSN 1367-4803. doi: 10.1093/bioinformatics/btag026. Fulltekst i vitenarkiv Vis sammendrag

  The accurate development, assessment, interpretation, and benchmarking of bioinformatics frameworks for analyzing transcriptional regulatory grammars rely on controlled simulations to validate the underlying methods. However, existing simulators often lack end-to-end flexibility or ease of integration, which limits their practical use. We present inMOTIFin, a lightweight, modular, and user-friendly Python-based software that addresses these gaps by providing versatile and efficient simulation and modification of DNA regulatory sequences. inMOTIFin enables users to simulate or modify regulatory sequences efficiently for the customizable generation of motifs and insertion of motif instances with precise control over their positions, co-occurrences, and spacing, as well as direct modification of real sequences, facilitating a comprehensive evaluation of motif-based methods and interpretation tools. We demonstrate inMOTIFin applications for the assessment of de novo motif discovery, the analysis of transcription factor cooperativity, and the support of explainability analyses for deep learning models. inMOTIFin ensures robust and reproducible analyses for studying transcriptional regulatory grammars.

• Kanduri, Chakravarthi; Mamica, Maria Anna; Olstad, Emilie Willoch; Zucknick, Manuela; Jessica, Li, Jingyi & Sandve, Geir Kjetil Ferkingstad (2025). Beware of counter-intuitive levels of false discoveries in datasets with strong intra-correlations. Genome Biology. ISSN 1465-6906. 26. doi: 10.1186/s13059-025-03734-z. Fulltekst i vitenarkiv Vis sammendrag

  The false discovery rate (FDR) controlling method by Benjamini and Hochberg (BH) is a popular choice in the omics fields. Here, we demonstrate that in datasets with a large degree of dependencies between features, FDR correction methods like BH can sometimes counter-intuitively report very high numbers of false positives, potentially misleading researchers. We call the attention of researchers to use suited multiple testing strategies and approaches like synthetic null data (negative control) to identify and minimize caveats related to false discoveries, as in the cases where false findings do occur, they may be numerous.

• Alwash, Mostafa; Hajj, Ghadi al; Grytten, Ivar & Sandve, Geir Kjetil Ferkingstad (2025). Meta simulation approach for evaluating machine learning method selection in data limited settings. Scientific Reports. 15. doi: 10.1038/s41598-025-24627-y. Fulltekst i vitenarkiv Vis sammendrag

  Selecting appropriate machine learning (ML) methods for domain-specific tasks remains a persistent challenge, particularly in medicine where datasets are often small, heterogeneous, and incomplete. Traditional benchmarking strategies rely on limited observational samples, which may not capture the complexity of the underlying data-generating process (DGP). As a result, methods that perform well on available data may generalise poorly in real-world practice. We present SimCalibration, a meta-simulation framework that leverages structural learners (SLs) to infer an approximated data-generating process from limited data and generate synthetic datasets for large-scale benchmarking. This framework enables systematic evaluation of machine learning method selection strategies in settings where the true data-generating process is either known or can be approximated, allowing both validation against the ground truth and the generation of synthetic observations inferred from sparse samples. In rare disease research for example, where patient cohorts are inherently small, causal relationships are often conceptualised as directed acyclic graphs (DAGs). In this work, such structures are approximated directly from observational data, extending the utility of small datasets by enabling investigators to benchmark ML methods in a controlled simulation setting before deploying them in practice. This reduces the risk of selecting models that generalise poorly and supports more reliable decision-making in sensitive healthcare contexts. Experiments demonstrate that (a) structural learners vary in their ability to recover representative simulations for benchmarking, (b) structural learner-based benchmarking reduces variance in performance estimates compared to traditional validation, and (c) in some cases, structural learner-based approaches yield rankings that more closely match true relative performance than those derived from limited datasets. These findings highlight the value of simulation-based benchmarking for domains where drawing generalisable conclusions is critical, such as medicine, and offer greater transparency into the assumptions underlying predictive decisions.

• Ursu, Eugen; Minnegalieva, Aygul; Rawat, Puneet; Chernigovskaya, Maria; Tacutu, Robi & Sandve, Geir Kjetil Ferkingstad [Vis alle 8 forfattere av denne artikkelen] (2025). Training data composition determines machine learning generalization and biological rule discovery. Nature Machine Intelligence. 7(8), s. 1206–1219. doi: 10.1038/s42256-025-01089-5. Fulltekst i vitenarkiv Vis sammendrag

  Supervised machine learning models depend on training datasets containing positive and negative examples: dataset composition directly impacts model performance and bias. Given the importance of machine learning for immunotherapeutic design, we examined how different negative class definitions affect model generalization and rule discovery for antibody–antigen binding. Using synthetic-structure-based binding data, we evaluated models trained with various definitions of negative sets. Our findings reveal that high out-of-distribution performance can be achieved when the negative dataset contains more similar samples to the positive dataset, despite lower in-distribution performance. Furthermore, by leveraging ground-truth information, we show that binding rules associated with positive data change based on the negative data used. Validation on experimental data supported simulation-based observations. This work underscores the role of dataset composition in creating robust, generalizable and biology-aware sequence-based ML models

• Chernigovskaya, Maria; Quy, Khang Le; Stensland, Maria; Singh, Sachin; Nelson, Rowan & Yilmaz, Melih [Vis alle 19 forfattere av denne artikkelen] (2025). Systematic benchmarking of mass spectrometry-based antibody sequencing reveals methodological biases. Cell Systems. ISSN 2405-4712. 16(11). doi: 10.1016/j.cels.2025.101449. Fulltekst i vitenarkiv Vis sammendrag

  The circulating antibody (Ab) repertoire is crucial for immune protection, holding significant immunological and biotechnological value. While bottom-up mass spectrometry (MS) is widely used for profiling the sequence diversity of circulating Abs (Ab repertoire sequencing [Ab-seq]), it has not been thoroughly benchmarked. We quantified the replicability and robustness of Ab-seq using six monoclonal Ab spike-ins in 70 combinations of concentration and oligoclonality, with and without polyclonal serum immunoglobulin G (IgG) background. Each combination underwent four protease treatments and was analyzed across four experimental and three technical replicates, totaling 3,360 liquid chromatography-tandem MS (LC-MS/MS) runs. We quantified the dependence of Ab-seq identification on Ab sequence, concentration, protease, presence of background IgGs, and bioinformatics methods. Integrating the data from experimental replicates, proteases, and bioinformatics tools enhanced Ab identification. De novo sequencing performed similarly to database-dependent methods at higher Ab concentrations, but de novo Ab reconstruction remains challenging. Our work provides a foundational resource for the field of MS-based Ab profiling. A record of this paper’s transparent peer review process is included in the supplemental information

• Riccardi, Enrico; Minotto, Thomas; Haff, Ingrid Hobæk & Sandve, Geir Kjetil Ferkingstad (2025). Connecting instance and model sparsity in multiple instance learning. Machine Learning with Applications (MLWA). 21, s. 100679–100679. doi: 10.1016/j.mlwa.2025.100679. Fulltekst i vitenarkiv
• Minotto, Thomas; Haff, Ingrid Hobæk; Riccardi, Enrico & Sandve, Geir Kjetil Ferkingstad (2025). Detecting statistical interactions in immune receptor data: a comparative study. Journal of Applied Statistics. ISSN 0266-4763. s. 22–22. doi: 10.1080/02664763.2025.2533483. Fulltekst i vitenarkiv
• Reddy, Akhila Srinivas; Fawad, Fatima; Leite, Mila Apostolova; Olstad, Emilie Willoch; Gervin, Kristina & Kanduri, Chakravarthi [Vis alle 11 forfattere av denne artikkelen] (2025). Prenatal exposure to psychotropics and analgesics on cognitive, linguistic and educational outcomes – a scoping review with focus on validity and reliability of outcome measures. BMC Pregnancy and Childbirth. 25. doi: 10.1186/s12884-025-07235-6. Fulltekst i vitenarkiv Vis sammendrag

  Objective To provide an overview of the observational studies on child’s cognitive, linguistic, and educational outcomes following prenatal exposure to psychotropics and analgesics, including reporting of outcome measure validity and reliability. Study design We searched four databases, MEDLINE, Embase, PsycINFO, and PubMed from inception to September 2023. We included all original studies involving participants less than 18 years old, who were prenatally exposed to psychotropics and/or analgesics with cognitive, linguistic, and/or educational outcomes and excluded those lacking comparison groups. Results 80 studies were identified. Most studies (47%) focused on the effects of prenatal exposure to antiepileptics on child cognition. Valproate was consistently associated with an increased risk of neurodevelopmental disorders, whereas the results for other medications were sparse and conflicting. Neurodevelopmental outcomes were measured by psychometric assessments in 71 studies and by diagnostic codes in health care registries in nine studies. Only 33 of the 71 studies (46.5%) using psychometric measures mentioned the psychometric properties of the instruments used. In studies using diagnostic outcome measures, only one study reported positive predictive values and performed a sensitivity analysis to address outcome misclassification. Conclusion Except for valproate, there is a concerning lack of studies on the impact of prenatal exposure to psychotropics and analgesics on cognitive, linguistic, and educational outcomes with existing studies yielding inconsistent findings. Regardless of whether psychometric measures or diagnostic codes were used, most studies lacked a robust assessment of outcome measures, which threatens their validity and interpretability. Future studies on long-term prenatal medication safety need to focus on the accuracy of neurodevelopmental outcome measures.

• Chernigovskaia, Maria; Pavlović, Milena; Kanduri, Chakravarthi; Gielis, Sofie; Robert, Philippe & Scheffer, Lonneke [Vis alle 12 forfattere av denne artikkelen] (2025). Simulation of adaptive immune receptors and repertoires with complex immune information to guide the development and benchmarking of AIRR machine learning. Nucleic Acids Research (NAR). ISSN 0305-1048. 53(3). doi: 10.1093/nar/gkaf025. Fulltekst i vitenarkiv Vis sammendrag

  Machine learning (ML) has shown great potential in the adaptive immune receptor repertoire (AIRR) field. However, there is a lack of large-scale ground-truth experimental AIRR data suitable for AIRR-ML-based disease diagnostics and therapeutics discovery. Simulated ground-truth AIRR data are required to complement the development and benchmarking of robust and interpretable AIRR-ML methods where experimental data is currently inaccessible or insufficient. The challenge for simulated data to be useful is incorporating key features observed in experimental repertoires. These features, such as antigen or disease-associated immune information, cause AIRR-ML problems to be challenging. Here, we introduce LIgO, a software suite, which simulates AIRR data for the development and benchmarking of AIRR-ML methods. LIgO incorporates different types of immune information both on the receptor and the repertoire level and preserves native-like generation probability distribution. Additionally, LIgO assists users in determining the computational feasibility of their simulations. We show two examples where LIgO supports the development and validation of AIRR-ML methods: (i) how individuals carrying out-of-distribution immune information impacts receptor-level prediction performance and (ii) how immune information co-occurring in the same AIRs impacts the performance of conventional receptor-level encoding and repertoire-level classification approaches. LIgO guides the advancement and assessment of interpretable AIRR-ML methods.

• Johansen, Anette; Sandve, Geir Kjetil Ferkingstad; Maxwell, Joseph R.; Smithson, Glennda; Sollid, Ludvig Magne & Stamnæs, Jorunn (2024). Biopsy Proteome Score Performs Well as an Effect Measure in a Gluten Challenge Trial of Celiac Disease. Clinical Gastroenterology and Hepatology. ISSN 1542-3565. 23(5), s. 758–765.e8. doi: 10.1016/j.cgh.2024.08.005. Fulltekst i vitenarkiv Vis sammendrag

  Background and Aims Development of novel treatments for celiac disease is dependent on precise tools to monitor changes in gluten-induced mucosal damage. Current histology measures are subjective and difficult to standardize. Biopsy proteome scoring is an objective alternative to histology which is based on robust changes in biological pathways that directly reflect gluten-induced mucosal damage. In this study, we aimed to evaluate biopsy proteome scoring as an effect measure in a clinical trial setting by measuring intestinal remodeling in response to oral gluten challenge. Methods We analyzed biopsies from a 14-day gluten challenge trial of treated celiac disease patients that consumed 3 g (n = 6) or 10 g (n = 7) gluten per day. Sections from individually embedded biopsies collected before and after challenge were processed for proteome scoring (n = 109) and measurement of villus height-to-crypt depth ratio (n = 58). Proteome scores were compared with histology, intraepithelial lymphocyte frequency and plasma interleukin-2. Results Change in proteome scores were significant for the group of patients who consumed 10 g gluten, but not for the group who consumed 3 g gluten. Altogether, 8 of 13 patients had changes in delta proteome scores above the cutoff. Proteome scores correlated with villus height-to-crypt depth ratios both at run-in and at day 15. Proteome scores at day 15 correlated with intraepithelial lymphocyte frequency and with plasma interleukin-2 levels measured 4 hours post–gluten intake. Conclusion Biopsy proteome scoring is a simple and reliable measure of gluten-induced mucosal remodeling in response to 14-day oral gluten challenge. ClinicalTrials.gov, Number: NCT03409796.

• Hajj, Ghadi al; Hubin, Aliaksandr; Kanduri, Chakravarthi; Pavlovic, Milena; Rand, Knut Dagestad & Widrich, Michael [Vis alle 11 forfattere av denne artikkelen] (2024). Incorporating probabilistic domain knowledge into deep multiple instance learning. Proceedings of Machine Learning Research (PMLR). 235, s. 17279–17297. Fulltekst i vitenarkiv Vis sammendrag

  Deep learning methods, including deep multiple instance learning methods, have been criticized for their limited ability to incorporate domain knowledge. A reason that knowledge incorporation is challenging in deep learning is that the models usually lack a mapping between their model components and the entities of the domain, making it a non-trivial task to incorporate probabilistic prior information. In this work, we show that such a mapping between domain entities and model components can be defined for a multiple instance learning setting and propose a framework DeeMILIP that encompasses multiple strategies to exploit this mapping for prior knowledge incorporation. We motivate and formalize these strategies from a probabilistic perspective. Experiments on an immune-based diagnostics case show that our proposed strategies allow to learn generalizable models even in settings with weak signals, limited dataset size, and limited compute.

• Pavlović, Milena; Hajj, Ghadi al; Kanduri, Chakravarthi; Pensar, Johan; Wood, Mollie Elizabeth & Sollid, Ludvig Magne [Vis alle 8 forfattere av denne artikkelen] (2024). Improving generalization of machine learning-identified biomarkers using causal modelling with examples from immune receptor diagnostics. Nature Machine Intelligence. 6(1), s. 15–24. doi: 10.1038/s42256-023-00781-8. Fulltekst i vitenarkiv
• Johansen, Anette; Sandve, Geir Kjetil Ferkingstad; Ibsen, Jostein Holen; Lundin, Knut; Sollid, Ludvig Magne & Stamnæs, Jorunn (2024). Biopsy Proteome Scoring to Determine Mucosal Remodeling in Celiac Disease. Gastroenterology. ISSN 0016-5085. 167(3), s. 493–504. doi: 10.1053/j.gastro.2024.03.006. Fulltekst i vitenarkiv Vis sammendrag

  Background & Aims Histologic evaluation of gut biopsies is a cornerstone for diagnosis and management of celiac disease (CeD). Despite its wide use, the method depends on proper biopsy orientation, and it suffers from interobserver variability. Biopsy proteome measurement reporting on the tissue state can be obtained by mass spectrometry analysis of formalin-fixed paraffin-embedded tissue. Here we aimed to transform biopsy proteome data into numerical scores that give observer-independent measures of mucosal remodeling in CeD. Methods A pipeline using glass-mounted formalin-fixed paraffin-embedded sections for mass spectrometry-based proteome analysis was established. Proteome data were converted to numerical scores using 2 complementary approaches: a rank-based enrichment score and a score based on machine learning using logistic regression. The 2 scoring approaches were compared with each other and with histology analyzing 18 patients with CeD with biopsies collected before and after treatment with a gluten-free diet as well as biopsies from patients with CeD with varying degree of remission (n = 22). Biopsies from individuals without CeD (n = 32) were also analyzed. Results The method yielded reliable proteome scoring of both unstained and H&E-stained glass-mounted sections. The scores of the 2 approaches were highly correlated, reflecting that both approaches pick up proteome changes in the same biological pathways. The proteome scores correlated with villus height-to-crypt depth ratio. Thus, the method is able to score biopsies with poor orientation. Conclusions Biopsy proteome scores give reliable observer and orientation-independent measures of mucosal remodeling in CeD. The proteomic method can readily be implemented by nonexpert laboratories in parallel to histology assessment and easily scaled for clinical trial settings.

• Reppe, Sjur; Gundersen, Sveinung; Sandve, Geir Kjetil Ferkingstad; Wang, Yunpeng; Andreassen, Ole A. & Medina-Gomez, Carolina [Vis alle 10 forfattere av denne artikkelen] (2024). Identification of Transcripts with Shared Roles in the Pathogenesis of Postmenopausal Osteoporosis and Cardiovascular Disease. International Journal of Molecular Sciences. ISSN 1661-6596. 25(10). doi: 10.3390/ijms25105554. Fulltekst i vitenarkiv Vis sammendrag

  Epidemiological evidence suggests existing comorbidity between postmenopausal osteoporosis (OP) and cardiovascular disease (CVD), but identification of possible shared genes is lacking. The skeletal global transcriptomes were analyzed in trans-iliac bone biopsies (n=84) from clinically well-characterized postmenopausal women (50 to 86 years) using microchips and RNA sequencing. Thousand transcripts highly correlated with areal bone mineral density (aBMD) were further analyzed using bioinformatics, and common genes overlapping with CVD and their biological mechanisms, pathways and functions were identified. Fifty genes (45 mRNAs, 5 miRNAs) were discovered with established roles in oxidative stress, inflammatory response, endothelial function, fibrosis, dyslipidemia and osteoblastogenesis/calcification. These pleiotropic genes with possible CVD comorbidity functions were also present in transcriptomes of microvascular endothelial cells and cardiomyocytes and differentially expressed between healthy and osteoporotic women with fragility fractures. The results were supported by a genetic pleiotropy-informed conditional False Discovery Rate approach identifying overlap in single nucleotide polymorphisms (SNPs) within several genes encoding the aBMD and CVD associated transcripts.The study provides transcriptional and genomic evidence for genes of importance for both BMD regulation and CVD risk in a large collection of postmenopausal bone biopsies. Most of the transcripts identified in the CVD risk categories have no previously recognized roles in OP pathogenesis and provide novel avenues for exploring the mechanistic basis for the biological association between CVD and OP.

• Vu, Mai Ha; Robert, Philippe; Akbar, Rahmad; Swiatczak, Bartlomiej; Sandve, Geir Kjetil Ferkingstad & Haug, Dag Trygve Truslew [Vis alle 7 forfattere av denne artikkelen] (2024). Linguistics-based formalization of the antibody language as a basis for antibody language models. Nature Computational Science. 4, s. 412–422. doi: 10.1038/s43588-024-00642-3. Fulltekst i vitenarkiv
• O'Donnell, Timothy J.; Kanduri, Chakravarthi; Isacchini, Giulio; Limenitakis, Julien P.; Brachman, Rebecca A. & Alvarez, Raymond A. [Vis alle 9 forfattere av denne artikkelen] (2024). Reading the repertoire: Progress in adaptive immune receptor analysis using machine learning. Cell Systems. ISSN 2405-4712. 15(12), s. 1168–1189. doi: 10.1016/j.cels.2024.11.006. Fulltekst i vitenarkiv Vis sammendrag

  The adaptive immune system holds invaluable information on past and present immune responses in the form of B and T cell receptor sequences, but we are limited in our ability to decode this information. Machine learning approaches are under active investigation for a range of tasks relevant to understanding and manipulating the adaptive immune receptor repertoire, including matching receptors to the antigens they bind, generating antibodies or T cell receptors for use as therapeutics, and diagnosing disease based on patient repertoires. Progress on these tasks has the potential to substantially improve the development of vaccines, therapeutics, and diagnostics, as well as advance our understanding of fundamental immunological principles. We outline key challenges for the field, highlighting the need for software benchmarking, targeted large-scale data generation, and coordinated research efforts.

• Bashour, Habib; Smorodina, Eva; Pariset, Matteo; Zhong, Jahn; Akbar, Rahmad & Chernigovskaia, Maria [Vis alle 15 forfattere av denne artikkelen] (2024). Biophysical cartography of the native and human-engineered antibody landscapes quantifies the plasticity of antibody developability. Communications Biology. 7(1), s. 1–25. doi: 10.1038/s42003-024-06561-3. Fulltekst i vitenarkiv Vis sammendrag

  Designing effective monoclonal antibody (mAb) therapeutics faces a multi-parameter optimization challenge known as “developability”, which reflects an antibody’s ability to progress through development stages based on its physicochemical properties. While natural antibodies may provide valuable guidance for mAb selection, we lack a comprehensive understanding of natural developability parameter (DP) plasticity (redundancy, predictability, sensitivity) and how the DP landscapes of human-engineered and natural antibodies relate to one another. These gaps hinder fundamental developability profile cartography. To chart natural and engineered DP landscapes, we computed 40 sequence- and 46 structure-based DPs of over two million native and human-engineered single-chain antibody sequences. We find lower redundancy among structure-based compared to sequence-based DPs. Sequence DP sensitivity to single amino acid substitutions varied by antibody region and DP, and structure DP values varied across the conformational ensemble of antibody structures. We show that sequence DPs are more predictable than structure-based ones across different machine-learning tasks and embeddings, indicating a constrained sequence-based design space. Human-engineered antibodies localize within the developability and sequence landscapes of natural antibodies, suggesting that human-engineered antibodies explore mere subspaces of the natural one. Our work quantifies the plasticity of antibody developability, providing a fundamental resource for multi-parameter therapeutic mAb design.

• Minotto, Thomas; Robert, Philippe; Haff, Ingrid Hobæk & Sandve, Geir Kjetil Ferkingstad (2024). Assessing the feasibility of statistical inference using synthetic antibody-antigen datasets. Statistical Applications in Genetics and Molecular Biology. 23:20230027(1), s. 1–14. doi: 10.1515/sagmb-2023-0027. Fulltekst i vitenarkiv Vis sammendrag

  Simulation frameworks are useful to stress-test predictive models when data is scarce, or to assert model sensitivity to specific data distributions. Such frameworks often need to recapitulate several layers of data complexity, including emergent properties that arise implicitly from the interaction between simulation components. Antibody-antigen binding is a complex mechanism by which an antibody sequence wraps itself around an antigen with high affinity. In this study, we use a synthetic simulation framework for antibody-antigen folding and binding on a 3D lattice that include full details on the spatial conformation of both molecules. We investigate how emergent properties arise in this framework, in particular the physical proximity of amino acids, their presence on the binding interface, or the binding status of a sequence, and relate that to the individual and pairwise contributions of amino acids in statistical models for binding prediction. We show that weights learnt from a simple logistic regression model align with some but not all features of amino acids involved in the binding, and that predictive sequence binding patterns can be enriched. In particular, main effects correlated with the capacity of a sequence to bind any antigen, while statistical interactions were related to sequence specificity.

• Hsieh, Ping-Han; Hsiao, Ru-Xiu; Ferenc, Katalin; Mathelier, Anthony; Burkholz, Rebekka & Chen, Chien-Yu [Vis alle 9 forfattere av denne artikkelen] (2024). CAVACHON: a hierarchical variational autoencoder to integrate multi-modal single-cell data. arXiv.org. ISSN 2331-8422. doi: 10.48550/arXiv.2405.18655. Fulltekst i vitenarkiv
• Scheffer, Lonneke; Reber, Eric Emanuel; Mehta, Brij Bhushan; Pavlović, Milena; Chernigovskaia, Maria & Richardson, Eve [Vis alle 11 forfattere av denne artikkelen] (2024). Predictability of antigen binding based on short motifs in the antibody CDRH3. Briefings in Bioinformatics. ISSN 1467-5463. 25(6), s. 1–10. doi: 10.1093/bib/bbae537. Fulltekst i vitenarkiv Vis sammendrag

  Adaptive immune receptors, such as antibodies and T-cell receptors, recognize foreign threats with exquisite specificity. A major challenge in adaptive immunology is discovering the rules governing immune receptor–antigen binding in order to predict the antigen binding status of previously unseen immune receptors. Many studies assume that the antigen binding status of an immune receptor may be determined by the presence of a short motif in the complementarity determining region 3 (CDR3), disregarding other amino acids. To test this assumption, we present a method to discover short motifs which show high precision in predicting antigen binding and generalize well to unseen simulated and experimental data. Our analysis of a mutagenesis-based antibody dataset reveals 11 336 position-specific, mostly gapped motifs of 3–5 amino acids that retain high precision on independently generated experimental data. Using a subset of only 178 motifs, a simple classifier was made that on the independently generated dataset outperformed a deep learning model proposed specifically for such datasets. In conclusion, our findings support the notion that for some antibodies, antigen binding may be largely determined by a short CDR3 motif. As more experimental data emerge, our methodology could serve as a foundation for in-depth investigations into antigen binding signals.

• Hsieh, Ping-Han; Lopes-Ramos, Camila Miranda; Zucknick, Manuela; Sandve, Geir Kjetil Ferkingstad; Glass, Kimberly & Kuijjer, Marieke (2023). Adjustment of spurious correlations in co-expression measurements from RNA-Sequencing data. Bioinformatics. ISSN 1367-4803. 39(10). doi: 10.1093/bioinformatics/btad610. Fulltekst i vitenarkiv Vis sammendrag

  Abstract Motivation Gene co-expression measurements are widely used in computational biology to identify coordinated expression patterns across a group of samples. Coordinated expression of genes may indicate that they are controlled by the same transcriptional regulatory program, or involved in common biological processes. Gene co-expression is generally estimated from RNA-Sequencing data, which are commonly normalized to remove technical variability. Here, we demonstrate that certain normalization methods, in particular quantile-based methods, can introduce false-positive associations between genes. These false-positive associations can consequently hamper downstream co-expression network analysis. Quantile-based normalization can, however, be extremely powerful. In particular, when preprocessing large-scale heterogeneous data, quantile-based normalization methods such as smooth quantile normalization can be applied to remove technical variability while maintaining global differences in expression for samples with different biological attributes. Results We developed SNAIL (Smooth-quantile Normalization Adaptation for the Inference of co-expression Links), a normalization method based on smooth quantile normalization specifically designed for modeling of co-expression measurements. We show that SNAIL avoids formation of false-positive associations in co-expression as well as in downstream network analyses. Using SNAIL, one can avoid arbitrary gene filtering and retain associations to genes that only express in small subgroups of samples. This highlights the method’s potential future impact on network modeling and other association-based approaches in large-scale heterogeneous data. Availability and implementation The implementation of the SNAIL algorithm and code to reproduce the analyses described in this work can be found in the GitHub repository https://github.com/kuijjerlab/PySNAIL.

• Kanduri, Chakravarthi; Scheffer, Lonneke; Pavlović, Milena; Rand, Knut Dagestad; Chernigovskaia, Maria & Pirvandy, Oz [Vis alle 9 forfattere av denne artikkelen] (2023). simAIRR: simulation of adaptive immune repertoires with realistic receptor sequence sharing for benchmarking of immune state prediction methods. GigaScience. 12, s. 1–16. doi: 10.1093/gigascience/giad074. Fulltekst i vitenarkiv Vis sammendrag

  Abstract Background Machine learning (ML) has gained significant attention for classifying immune states in adaptive immune receptor repertoires (AIRRs) to support the advancement of immunodiagnostics and therapeutics. Simulated data are crucial for the rigorous benchmarking of AIRR-ML methods. Existing approaches to generating synthetic benchmarking datasets result in the generation of naive repertoires missing the key feature of many shared receptor sequences (selected for common antigens) found in antigen-experienced repertoires. Results We demonstrate that a common approach to generating simulated AIRR benchmark datasets can introduce biases, which may be exploited for undesired shortcut learning by certain ML methods. To mitigate undesirable access to true signals in simulated AIRR datasets, we devised a simulation strategy (simAIRR) that constructs antigen-experienced-like repertoires with a realistic overlap of receptor sequences. simAIRR can be used for constructing AIRR-level benchmarks based on a range of assumptions (or experimental data sources) for what constitutes receptor-level immune signals. This includes the possibility of making or not making any prior assumptions regarding the similarity or commonality of immune state–associated sequences that will be used as true signals. We demonstrate the real-world realism of our proposed simulation approach by showing that basic ML strategies perform similarly on simAIRR-generated and real-world experimental AIRR datasets. Conclusions This study sheds light on the potential shortcut learning opportunities for ML methods that can arise with the state-of-the-art way of simulating AIRR datasets. simAIRR is available as a Python package: https://github.com/KanduriC/simAIRR.

• Olstad, Emilie Willoch; Nordeng, Hedvig Marie Egeland; Sandve, Geir Kjetil Ferkingstad; Lyle, Robert & Gervin, Kristina (2023). Effects of prenatal exposure to (es)citalopram and maternal depression during pregnancy on DNA methylation and child neurodevelopment. Translational Psychiatry. 13(1). doi: 10.1038/s41398-023-02441-2. Fulltekst i vitenarkiv Vis sammendrag

  Abstract Studies assessing associations between prenatal exposure to antidepressants, maternal depression, and offspring DNA methylation (DNAm) have been inconsistent. Here, we investigated whether prenatal exposure to citalopram or escitalopram ((es)citalopram) and maternal depression is associated with differences in DNAm. Then, we examined if there is an interaction effect of (es)citalopram exposure and DNAm on offspring neurodevelopmental outcomes. Finally, we investigated whether DNAm at birth correlates with neurodevelopmental trajectories in childhood. We analyzed DNAm in cord blood from the Norwegian Mother, Father and Child Cohort Study (MoBa) biobank. MoBa contains questionnaire data on maternal (es)citalopram use and depression during pregnancy and information about child neurodevelopmental outcomes assessed by internationally recognized psychometric tests. In addition, we retrieved ADHD diagnoses from the Norwegian Patient Registry and information on pregnancies from the Medical Birth Registry of Norway. In total, 958 newborn cord blood samples were divided into three groups: (1) prenatal (es)citalopram exposed ( n  = 306), (2) prenatal maternal depression exposed ( n  = 308), and (3) propensity score-selected controls ( n  = 344). Among children exposed to (es)citalopram, there were more ADHD diagnoses and symptoms and delayed communication and psychomotor development. We did not identify differential DNAm associated with (es)citalopram or depression, nor any interaction effects on neurodevelopmental outcomes throughout childhood. Trajectory modeling identified subgroups of children following similar developmental patterns. Some of these subgroups were enriched for children exposed to maternal depression, and some subgroups were associated with differences in DNAm at birth. Interestingly, several of the differentially methylated genes are involved in neuronal processes and development. These results suggest DNAm as a potential predictive molecular marker of later abnormal neurodevelopmental outcomes, but we cannot conclude whether DNAm links prenatal (es)citalopram exposure or maternal depression with child neurodevelopmental outcomes.

• Kalyanasundaram, Sumana; Lefol, Yohan Pierre; Gundersen, Sveinung; Rognes, Torbjørn; Alsøe, Lene & Nilsen, Hilde [Vis alle 9 forfattere av denne artikkelen] (2023). hGSuite HyperBrowser: A web-based toolkit for hierarchical metadata-informed analysis of genomic tracks. PLOS ONE. 18(7). doi: 10.1371/journal.pone.0286330. Fulltekst i vitenarkiv Vis sammendrag

  Many high-throughput sequencing datasets can be represented as objects with coordinates along a reference genome. Currently, biological investigations often involve a large number of such datasets, for example representing different cell types or epigenetic factors. Drawing overall conclusions from a large collection of results for individual datasets may be challenging and time-consuming. Meaningful interpretation often requires the results to be aggregated according to metadata that represents biological characteristics of interest. In this light, we here propose the hierarchical Genomic Suite HyperBrowser (hGSuite), an open-source extension to the GSuite HyperBrowser platform, which aims to provide a means for extracting key results from an aggregated collection of high-throughput DNA sequencing data. The hGSuite utilizes a metadata-informed data cube to calculate various statistics across the multiple dimensions of the datasets. With this work, we show that the hGSuite and its associated data cube methodology offers a quick and accessible way for exploratory analysis of large genomic datasets. The web-based toolkit named hGsuite Hyperbrowser is available at https://hyperbrowser.uio.no/hgsuite under a GPLv3 license.

• Shakibfar, Saeed; Nyberg, Fredrik; Li, Huiqi; Zhao, Jing; Nordeng, Hedvig Marie Egeland & Sandve, Geir Kjetil Ferkingstad [Vis alle 10 forfattere av denne artikkelen] (2023). Artificial intelligence-driven prediction of COVID-19-related hospitalization and death: a systematic review. Frontiers in Public Health. 11. doi: 10.3389/fpubh.2023.1183725. Fulltekst i vitenarkiv Vis sammendrag

  Aim To perform a systematic review on the use of Artificial Intelligence (AI) techniques for predicting COVID-19 hospitalization and mortality using primary and secondary data sources. Study eligibility criteria Cohort, clinical trials, meta-analyses, and observational studies investigating COVID-19 hospitalization or mortality using artificial intelligence techniques were eligible. Articles without a full text available in the English language were excluded. Data sources Articles recorded in Ovid MEDLINE from 01/01/2019 to 22/08/2022 were screened. Data extraction We extracted information on data sources, AI models, and epidemiological aspects of retrieved studies. Bias assessment A bias assessment of AI models was done using PROBAST. Participants Patients tested positive for COVID-19. Results We included 39 studies related to AI-based prediction of hospitalization and death related to COVID-19. The articles were published in the period 2019-2022, and mostly used Random Forest as the model with the best performance. AI models were trained using cohorts of individuals sampled from populations of European and non-European countries, mostly with cohort sample size <5,000. Data collection generally included information on demographics, clinical records, laboratory results, and pharmacological treatments (i.e., high-dimensional datasets). In most studies, the models were internally validated with cross-validation, but the majority of studies lacked external validation and calibration. Covariates were not prioritized using ensemble approaches in most of the studies, however, models still showed moderately good performances with Area under the Receiver operating characteristic Curve (AUC) values >0.7. According to the assessment with PROBAST, all models had a high risk of bias and/or concern regarding applicability. Conclusions A broad range of AI techniques have been used to predict COVID-19 hospitalization and mortality. The studies reported good prediction performance of AI models, however, high risk of bias and/or concern regarding applicability were detected.

• Chlubnova, Marketa; Christophersen, Asbjørn Otto; Sandve, Geir Kjetil Ferkingstad; Lundin, Knut Erik Aslaksen; Jahnsen, Jørgen & Dahal-Koirala, Shiva [Vis alle 7 forfattere av denne artikkelen] (2023). Identification of gluten T cell epitopes driving celiac disease. Science Advances. 9(4), s. 1–8. doi: 10.1126/sciadv.ade5800. Fulltekst i vitenarkiv Vis sammendrag

  CD4+ T cells specific for cereal gluten proteins are key players in celiac disease (CeD) pathogenesis. While several CeD-relevant gluten T cell epitopes have been identified, epitopes recognized by a substantial proportion of gluten-reactive T cells remain unknown. The identification of such CeD-driving gluten epitopes is important for the food industry and in clinical settings. Here, we have combined the knowledge of a distinct phenotype of gluten-reactive T cells and key features of known gluten epitopes for the discovery of unknown epitopes. We tested 42 wheat gluten–reactive T cell clones, isolated on the basis of their distinct phenotype and with no reactivity to known epitopes, against a panel of synthetic peptides bioinformatically identified from a wheat gluten protein database. We were able to assign reactivity to 10 T cell clones and identified a 9-nucleotide oligomer core region of five previously uncharacterized gliadin/glutenin epitopes. This work represents an advance in the effort to identify CeD-driving gluten epitopes.

• Hajj, Ghadi al; Pensar, Johan & Sandve, Geir Kjetil Ferkingstad (2023). DagSim: Combining DAG-based model structure with unconstrained data types and relations for flexible, transparent, and modularized data simulation. PLOS ONE. 18(4). doi: 10.1371/journal.pone.0284443. Fulltekst i vitenarkiv Vis sammendrag

  Data simulation is fundamental for machine learning and causal inference, as it allows exploration of scenarios and assessment of methods in settings with full control of ground truth. Directed acyclic graphs (DAGs) are well established for encoding the dependence structure over a collection of variables in both inference and simulation settings. However, while modern machine learning is applied to data of an increasingly complex nature, DAG-based simulation frameworks are still confined to settings with relatively simple variable types and functional forms. We here present DagSim, a Python-based framework for DAG-based data simulation without any constraints on variable types or functional relations. A succinct YAML format for defining the simulation model structure promotes transparency, while separate user-provided functions for generating each variable based on its parents ensure simulation code modularization. We illustrate the capabilities of DagSim through use cases where metadata variables control shapes in an image and patterns in bio-sequences. DagSim is available as a Python package at PyPI. Source code and documentation are available at: https://github.com/uio-bmi/dagsim

• Shakibfar, Saeed; Zhao, Jing; Li, Huiqi; Nordeng, Hedvig Marie Egeland; Lupattelli, Angela & Pavlović, Milena [Vis alle 12 forfattere av denne artikkelen] (2023). Machine learning-driven development of a disease risk score for COVID-19 hospitalization and mortality: a Swedish and Norwegian register-based study. Frontiers in Public Health. 11. doi: 10.3389/fpubh.2023.1258840. Fulltekst i vitenarkiv Vis sammendrag

  Aims To develop a disease risk score for COVID-19-related hospitalization and mortality in Sweden and externally validate it in Norway. Method We employed linked data from the national health registries of Sweden and Norway to conduct our study. We focused on individuals in Sweden with confirmed SARS-CoV-2 infection through RT-PCR testing up to August 2022 as our study cohort. Within this group, we identified hospitalized cases as those who were admitted to the hospital within 14 days of testing positive for SARS-CoV-2 and matched them with five controls from the same cohort who were not hospitalized due to SARS-CoV-2. Additionally, we identified individuals who died within 30 days after being hospitalized for COVID-19. To develop our disease risk scores, we considered various factors, including demographics, infectious, somatic, and mental health conditions, recorded diagnoses, and pharmacological treatments. We also conducted age-specific analyses and assessed model performance through 5-fold cross-validation. Finally, we performed external validation using data from the Norwegian population with COVID-19 up to December 2021. Results During the study period, a total of 124,560 individuals in Sweden were hospitalized, and 15,877 individuals died within 30 days following COVID-19 hospitalization. Disease risk scores for both hospitalization and mortality demonstrated predictive capabilities with ROC-AUC values of 0.70 and 0.72, respectively, across the entire study period. Notably, these scores exhibited a positive correlation with the likelihood of hospitalization or death. In the external validation using data from the Norwegian COVID-19 population (consisting of 53,744 individuals), the disease risk score predicted hospitalization with an AUC of 0.47 and death with an AUC of 0.74. Conclusion The disease risk score showed moderately good performance to predict COVID-19-related mortality but performed poorly in predicting hospitalization when externally validated.

• Wæhler, Hallvard Austin; Labba, Nils-Anders Johannes; Paulsen, Ragnhild Elisabeth Heimtun; Sandve, Geir Kjetil Ferkingstad & Eskeland, Ragnhild (2023). ANDA: an open-source tool for automated image analysis of in vitro neuronal cells. BMC Neuroscience. 24(1), s. 1–11. doi: 10.1186/s12868-023-00826-z. Fulltekst i vitenarkiv Vis sammendrag

  Background Imaging of in vitro neuronal differentiation and measurements of cell morphologies have led to novel insights into neuronal development. Live-cell imaging techniques and large datasets of images have increased the demand for automated pipelines for quantitative analysis of neuronal morphological metrics. Results ANDA is an analysis workflow that quantifies various aspects of neuronal morphology from high-throughput live-cell imaging screens of in vitro neuronal cell types. This tool automates the analysis of neuronal cell numbers, neurite lengths and neurite attachment points. We used chicken, rat, mouse, and human in vitro models for neuronal differentiation and have demonstrated the accuracy, versatility, and efficiency of the tool. Conclusions ANDA is an open-source tool that is easy to use and capable of automated processing from time-course measurements of neuronal cells. The strength of this pipeline is the capability to analyse high-throughput imaging screens.

• Minotto, Thomas; Haff, Ingrid Hobæk & Sandve, Geir Kjetil Ferkingstad (2022). Detecting statistical interactions in immune receptor datasets. I Torelli, Nicola; BELLIO, RUGGERO & MUGGEO, VITO (Red.), Proceedings of the 36th International Workshop on Statistical Modelling. EUT Edizioni Università di Trieste. ISSN 9788855113090. s. 253–257. Fulltekst i vitenarkiv Vis sammendrag

  Recent progress in the understanding of immune receptors suggests that complex interactions between amino acids are important in determining binding to antigens. Yet, current methods focus mostly on constructing good predictors for the data with complex models, and less on the understanding of the underlying interaction effects. Here, we attempt to retrieve high-order statistical interactions in immune receptor data with different methods. We study performance at this task in a large simulation study, and how it depends on the order, amount and strength of the interactions, witness rate and sample size. The results show that pairwise interactions are easily retrieved, but model complexity harms detection. Interactions are better detected in larger samples, but the process is then slower.

• Robert, Philippe Paul Auguste; Akbar, Rahmad; Frank, Robert; Pavlović, Milena; Widrich, Michael & Snapkov, Igor [Vis alle 25 forfattere av denne artikkelen] (2022). Unconstrained generation of synthetic antibody–antigen structures to guide machine learning methodology for antibody specificity prediction. Nature Computational Science. 2(12), s. 845–865. doi: 10.1038/s43588-022-00372-4. Fulltekst i vitenarkiv
• Dahal-Koirala, Shiva; Balaban, Gabriel; Neumann, Ralf Stefan; Scheffer, Lonneke; Lundin, Knut & Greiff, Victor [Vis alle 9 forfattere av denne artikkelen] (2022). TCRpower: quantifying the detection power of T-cell receptor sequencing with a novel computational pipeline calibrated by spike-in sequences. Briefings in Bioinformatics. ISSN 1467-5463. 23(2), s. 1–14. doi: 10.1093/bib/bbab566. Fulltekst i vitenarkiv Vis sammendrag

  T-cell receptor (TCR) sequencing has enabled the development of innovative diagnostic tests for cancers, autoimmune diseases and other applications. However, the rarity of many T-cell clonotypes presents a detection challenge, which may lead to misdiagnosis if diagnostically relevant TCRs remain undetected. To address this issue, we developed TCRpower, a novel computational pipeline for quantifying the statistical detection power of TCR sequencing methods. TCRpower calculates the probability of detecting a TCR sequence as a function of several key parameters: in-vivo TCR frequency, T-cell sample count, read sequencing depth and read cutoff. To calibrate TCRpower, we selected unique TCRs of 45 T-cell clones (TCCs) as spike-in TCRs. We sequenced the spike-in TCRs from TCCs, together with TCRs from peripheral blood, using a 5′ RACE protocol. The 45 spike-in TCRs covered a wide range of sample frequencies, ranging from 5 per 100 to 1 per 1 million. The resulting spike-in TCR read counts and ground truth frequencies allowed us to calibrate TCRpower. In our TCR sequencing data, we observed a consistent linear relationship between sample and sequencing read frequencies. We were also able to reliably detect spike-in TCRs with frequencies as low as one per million. By implementing an optimized read cutoff, we eliminated most of the falsely detected sequences in our data (TCR α-chain 99.0% and TCR β-chain 92.4%), thereby improving diagnostic specificity. TCRpower is publicly available and can be used to optimize future TCR sequencing experiments, and thereby enable reliable detection of disease-relevant TCRs for diagnostic applications.

• Kanduri, Chakravarthi; Pavlović, Milena; Scheffer, Lonneke; Motwani, Keshav; Chernigovskaia, Maria & Greiff, Victor [Vis alle 7 forfattere av denne artikkelen] (2022). Profiling the baseline performance and limits of machine learning models for adaptive immune receptor repertoire classification. GigaScience. 11(05). doi: 10.1093/gigascience/giac046. Fulltekst i vitenarkiv Vis sammendrag

  Background Machine learning (ML) methodology development for the classification of immune states in adaptive immune receptor repertoires (AIRRs) has seen a recent surge of interest. However, so far, there does not exist a systematic evaluation of scenarios where classical ML methods (such as penalized logistic regression) already perform adequately for AIRR classification. This hinders investigative reorientation to those scenarios where method development of more sophisticated ML approaches may be required. Results To identify those scenarios where a baseline ML method is able to perform well for AIRR classification, we generated a collection of synthetic AIRR benchmark data sets encompassing a wide range of data set architecture-associated and immune state–associated sequence patterns (signal) complexity. We trained ≈1,700 ML models with varying assumptions regarding immune signal on ≈1,000 data sets with a total of ≈250,000 AIRRs containing ≈46 billion TCRβ CDR3 amino acid sequences, thereby surpassing the sample sizes of current state-of-the-art AIRR-ML setups by two orders of magnitude. We found that L1-penalized logistic regression achieved high prediction accuracy even when the immune signal occurs only in 1 out of 50,000 AIR sequences. Conclusions We provide a reference benchmark to guide new AIRR-ML classification methodology by (i) identifying those scenarios characterized by immune signal and data set complexity, where baseline methods already achieve high prediction accuracy, and (ii) facilitating realistic expectations of the performance of AIRR-ML models given training data set properties and assumptions. Our study serves as a template for defining specialized AIRR benchmark data sets for comprehensive benchmarking of AIRR-ML methods.

• Olstad, Emilie Willoch; Nordeng, Hedvig Marie Egeland; Sandve, Geir Kjetil Ferkingstad; Lyle, Robert & Gervin, Kristina (2022). Low reliability of DNA methylation across Illumina Infinium platforms in cord blood: implications for replication studies and meta-analyses of prenatal exposures. Clinical Epigenetics. 14(1). doi: 10.1186/s13148-022-01299-3. Fulltekst i vitenarkiv Vis sammendrag

  Background: There is an increasing interest in the role of epigenetics in epidemiology, but the emerging research feld faces several critical biological and technical challenges. In particular, recent studies have shown poor correlation of measured DNA methylation (DNAm) levels within and across Illumina Infnium platforms in various tissues. In this study, we have investigated concordance between 450 k and EPIC Infnium platforms in cord blood. We could not replicate our previous fndings on the association of prenatal paracetamol exposure with cord blood DNAm, which prompted an investigation of cross-platform DNAm diferences. Results: This study is based on two DNAm data sets from cord blood samples selected from the Norwegian Mother, Father and Child Cohort Study (MoBa). DNAm of one data set was measured using the 450 k platform and the other data set was measured using the EPIC platform. Initial analyses of the EPIC data could not replicate any of our previous signifcant fndings in the 450 k data on associations between prenatal paracetamol exposure and cord blood DNAm. A subset of the samples (n=17) was included in both data sets, which enabled analyses of technical sources potentially contributing to the negative replication. Analyses of these 17 samples with repeated measurements revealed high per-sample correlations (−R ≈0.99), but low per-CpG correlations (−R≈0.24) between the platforms. 1.7% of the CpGs exhibited a mean DNAm diference across platforms >0.1. Furthermore, only 26.7% of the CpGs exhibited a moderate or better cross-platform reliability (intra-class correlation coefcient ≥0.5). Conclusion: The observations of low cross-platform probe correlation and reliability corroborate previous reports in other tissues. Our study cannot determine the origin of the diferences between platforms. Nevertheless, it emulates the setting in studies using data from multiple Infnium platforms, often analysed several years apart. Therefore, the fndings may have important implications for future epigenome-wide association studies (EWASs), in replication, meta-analyses and longitudinal studies. Cognisance and transparency of the challenges related to cross-platform studies may enhance the interpretation, replicability and validity of EWAS results both in cord blood and other tissues, ultimately improving the clinical relevance of epigenetic epidemiology.

• Rognes, Torbjørn; Scheffer, Lonneke; Greiff, Victor & Sandve, Geir Kjetil (2022). CompAIRR: ultra-fast comparison of adaptive immune receptor repertoires by exact and approximate sequence matching. Bioinformatics. ISSN 1367-4803. 38(17), s. 4230–4232. doi: 10.1093/bioinformatics/btac505. Fulltekst i vitenarkiv Vis sammendrag

  Abstract Motivation Adaptive immune receptor (AIR) repertoires (AIRRs) record past immune encounters with exquisite specificity. Therefore, identifying identical or similar AIR sequences across individuals is a key step in AIRR analysis for revealing convergent immune response patterns that may be exploited for diagnostics and therapy. Existing methods for quantifying AIRR overlap scale poorly with increasing dataset numbers and sizes. To address this limitation, we developed CompAIRR, which enables ultra-fast computation of AIRR overlap, based on either exact or approximate sequence matching. Results CompAIRR improves computational speed 1000-fold relative to the state of the art and uses only one-third of the memory: on the same machine, the exact pairwise AIRR overlap of 104 AIRRs with 105 sequences is found in ∼17 min, while the fastest alternative tool requires 10 days. CompAIRR has been integrated with the machine learning ecosystem immuneML to speed up commonly used AIRR-based machine learning applications. Availability and implementation CompAIRR code and documentation are available at https://github.com/uio-bmi/compairr. Docker images are available at https://hub.docker.com/r/torognes/compairr. The code to replicate the synthetic datasets, scripts for benchmarking and creating figures, and all raw data underlying the figures are available at https://github.com/uio-bmi/compairr-benchmarking. Supplementary information Supplementary data are available at Bioinformatics online.

• Grytten, Ivar; Rand, Knut Dagestad & Sandve, Geir Kjetil Ferkingstad (2022). KAGE: fast alignment-free graph-based genotyping of SNPs and short indels. Genome Biology. ISSN 1465-6906. 23(1). doi: 10.1186/s13059-022-02771-2. Fulltekst i vitenarkiv Vis sammendrag

  Genotyping is a core application of high-throughput sequencing. We present KAGE, a genotyper for SNPs and short indels that is inspired by recent developments within graph-based genome representations and alignment-free methods. KAGE uses a pan-genome representation of the population to efficiently and accurately predict genotypes. Two novel ideas improve both the speed and accuracy: a Bayesian model incorporates genotypes from thousands of individuals to improve prediction accuracy, and a computationally efficient method leverages correlation between variants. We show that the accuracy of KAGE is at par with the best existing alignment-free genotypers, while being an order of magnitude faster.

• Weber, Cédric R.; Rubio, Teresa; Wang, Longlong; Zhang, Wei; Robert, Philippe Paul Auguste & Akbar, Rahmad [Vis alle 15 forfattere av denne artikkelen] (2022). Reference-based comparison of adaptive immune receptor repertoires. Cell Reports Methods. 2(8). doi: 10.1016/j.crmeth.2022.100269. Fulltekst i vitenarkiv Vis sammendrag

  B and T cell receptor (immune) repertoires can represent an individual’s immune history. While current repertoire analysis methods aim to discriminate between health and disease states, they are typically based on only a limited number of parameters. Here, we introduce immuneREF: a quantitative multidimensional measure of adaptive immune repertoire (and transcriptome) similarity that allows interpretation of immune repertoire variation by relying on both repertoire features and cross-referencing of simulated and experimental datasets. To quantify immune repertoire similarity landscapes across health and disease, we applied immuneREF to >2,400 datasets from individuals with varying immune states (healthy, [autoimmune] disease, and infection). We discovered, in contrast to the current paradigm, that blood-derived immune repertoires of healthy and diseased individuals are highly similar for certain immune states, suggesting that repertoire changes to immune perturbations are less pronounced than previously thought. In conclusion, immuneREF enables the population-wide study of adaptive immune response similarity across immune states.

• Akbar, Rahmad; Robert, Philippe Paul Auguste; Weber, Cédric R.; Widrich, Michael; Frank, Robert & Pavlovic, Milena [Vis alle 19 forfattere av denne artikkelen] (2022). In silico proof of principle of machine learning-based antibody design at unconstrained scale. mAbs. ISSN 1942-0862. 14:e2031482(1), s. 1–18. doi: 10.1080/19420862.2022.2031482. Fulltekst i vitenarkiv Vis sammendrag

  Generative machine learning (ML) has been postulated to become a major driver in the computational design of antigen-specific monoclonal antibodies (mAb). However, efforts to confirm this hypothesis have been hindered by the infeasibility of testing arbitrarily large numbers of antibody sequences for their most critical design parameters: paratope, epitope, affinity, and developability. To address this challenge, we leveraged a lattice-based antibody-antigen binding simulation framework, which incorporates a wide range of physiological antibody-binding parameters. The simulation framework enables the computation of synthetic antibody-antigen 3D-structures, and it functions as an oracle for unrestricted prospective evaluation and benchmarking of antibody design parameters of ML-generated antibody sequences. We found that a deep generative model, trained exclusively on antibody sequence (one dimensional: 1D) data can be used to design conformational (three dimensional: 3D) epitope-specific antibodies, matching, or exceeding the training dataset in affinity and developability parameter value variety. Furthermore, we established a lower threshold of sequence diversity necessary for high-accuracy generative antibody ML and demonstrated that this lower threshold also holds on experimental real-world data. Finally, we show that transfer learning enables the generation of high-affinity antibody sequences from low-N training data. Our work establishes a priori feasibility and the theoretical foundation of high-throughput ML-based mAb design.

• Yao, Ying; Zia, Asima; Neumann, Ralf Stefan; Pavlović, Milena; Balaban, Gabriel & Lundin, Knut Erik Aslaksen [Vis alle 8 forfattere av denne artikkelen] (2021). T cell receptor repertoire as a potential diagnostic marker for celiac disease. Clinical Immunology. ISSN 1521-6616. 222:108621, s. 1–7. doi: 10.1016/j.clim.2020.108621. Fulltekst i vitenarkiv Vis sammendrag

  An individual's T cell repertoire is skewed towards some specificities as a result of past antigen exposure and subsequent clonal expansion. Identifying T cell receptor signatures associated with a disease is challenging due to the overall complexity of antigens and polymorphic HLA allotypes. In celiac disease, the antigen epitopes are well characterised and the specific HLA-DQ2-restricted T-cell repertoire associated with the disease has been explored in depth. By investigating T cell receptor repertoires of unsorted lamina propria T cells from 15 individuals, we provide the first proof-of-concept study showing that it could be possible to infer disease state by matching against a priori known disease-associated T cell receptor sequences.

• Slabodkin, Andrei; Chernigovskaia, Maria; Mikocziova, Ivana; Akbar, Rahmad; Scheffer, Lonneke & Pavlović, Milena [Vis alle 16 forfattere av denne artikkelen] (2021). Individualized VDJ recombination predisposes the available Ig sequence space. Genome Research. ISSN 1088-9051. 31(12), s. 2209–2225. doi: 10.1101/gr.275373.121. Fulltekst i vitenarkiv Vis sammendrag

  The process of recombination between variable (V), diversity (D), and joining (J) immunoglobulin (Ig) gene segments determines an individual's naive Ig repertoire and, consequently, (auto)antigen recognition. VDJ recombination follows probabilistic rules that can be modeled statistically. So far, it remains unknown whether VDJ recombination rules differ between individuals. If these rules differed, identical (auto)antigen-specific Ig sequences would be generated with individual-specific probabilities, signifying that the available Ig sequence space is individual specific. We devised a sensitivity-tested distance measure that enables inter-individual comparison of VDJ recombination models. We discovered, accounting for several sources of noise as well as allelic variation in Ig sequencing data, that not only unrelated individuals but also human monozygotic twins and even inbred mice possess statistically distinguishable immunoglobulin recombination models. This suggests that, in addition to genetic, there is also nongenetic modulation of VDJ recombination. We demonstrate that population-wide individualized VDJ recombination can result in orders of magnitude of difference in the probability to generate (auto)antigen-specific Ig sequences. Our findings have implications for immune receptor–based individualized medicine approaches relevant to vaccination, infection, and autoimmunity.

• Arnaout, Ramy A.; Prak, Eline T. Luning; Schwab, Nicholas; Rubelt, Florian; Greiff, Victor & Pavlović, Milena [Vis alle 32 forfattere av denne artikkelen] (2021). The Future of Blood Testing Is the Immunome. Frontiers in Immunology. 12:626793, s. 1–6. doi: 10.3389/fimmu.2021.626793. Fulltekst i vitenarkiv Vis sammendrag

  It is increasingly clear that an extraordinarily diverse range of clinically important conditions—including infections, vaccinations, autoimmune diseases, transplants, transfusion reactions, aging, and cancers—leave telltale signatures in the millions of V(D)J-rearranged antibody and T cell receptor [TR per the Human Genome Organization (HUGO) nomenclature but more commonly known as TCR] genes collectively expressed by a person’s B cells (antibodies) and T cells. We refer to these as the immunome . Because of its diversity and complexity, the immunome provides singular opportunities for advancing personalized medicine by serving as the substrate for a highly multiplexed, near-universal blood test. Here we discuss some of these opportunities, the current state of immunome-based diagnostics, and highlight some of the challenges involved. We conclude with a call to clinicians, researchers, and others to join efforts with the Adaptive Immune Receptor Repertoire Community (AIRR-C) to realize the diagnostic potential of the immunome.

• Yao, Ying; Wyrozemski, Lukasz Adam; Lundin, Knut; Sandve, Geir Kjetil Ferkingstad & Qiao, Shuo Wang (2021). Differential expression profile of gluten-specific T cells identified by single-cell RNA-seq. PLOS ONE. 16:e0258029(10), s. 1–17. doi: 10.1371/journal.pone.0258029. Fulltekst i vitenarkiv Vis sammendrag

  Gluten-specific CD4 + T cells drive the pathogenesis of celiac disease and circulating gluten-specific T cells can be identified by staining with HLA-DQ:gluten tetramers. In this first single-cell RNA-seq study of tetramer-sorted T cells from untreated celiac disease patients blood, we found that gluten-specific T cells showed distinct transcriptomic profiles consistent with activated effector memory T cells that shared features with Th1 and follicular helper T cells. Compared to non-specific cells, gluten-specific T cells showed differential expression of several genes involved in T-cell receptor signaling, translational processes, apoptosis, fatty acid transport, and redox potentials. Many of the gluten-specific T cells studied shared T-cell receptor with each other, indicating that circulating gluten-specific T cells belong to a limited number of clones. Moreover, the transcriptional profiles of cells that shared the same clonal origin were transcriptionally more similar compared with between clonally unrelated gluten-specific cells.

• Pavlović, Milena; Scheffer, Lonneke; Motwani, Keshav; Kanduri, Chakravarthi; Kompova, Radmila & Vazov, Nikolay Aleksandrov [Vis alle 42 forfattere av denne artikkelen] (2021). The immuneML ecosystem for machine learning analysis of adaptive immune receptor repertoires. Nature Machine Intelligence. 3(11), s. 936–944. doi: 10.1038/s42256-021-00413-z. Fulltekst i vitenarkiv
• Dahal-Koirala, Shiva; Risnes, Louise Fremgaard; Neumann, Ralf Stefan; Christophersen, Asbjørn; Lundin, Knut E. A. & Sandve, Geir Kjetil [Vis alle 8 forfattere av denne artikkelen] (2021). Comprehensive Analysis of CDR3 Sequences in Gluten-Specific T-Cell Receptors Reveals a Dominant R-Motif and Several New Minor Motifs. Frontiers in Immunology. 12:639672, s. 1–13. doi: 10.3389/fimmu.2021.639672. Fulltekst i vitenarkiv Vis sammendrag

  Gluten-specific CD4+ T cells are drivers of celiac disease (CeD). Previous studies of gluten-specific T-cell receptor (TCR) repertoires have found public TCRs shared across multiple individuals, biased usage of particular V-genes and conserved CDR3 motifs. The CDR3 motifs within the gluten-specific TCR repertoire, however, have not been systematically investigated. In the current study, we analyzed the largest TCR database of gluten-specific CD4+ T cells studied so far consisting of TCRs of 3122 clonotypes from 63 CeD patients. We established a TCR database from CD4+ T cells isolated with a mix of HLA-DQ2.5:gluten tetramers representing four immunodominant gluten epitopes. In an unbiased fashion we searched by hierarchical clustering for common CDR3 motifs among 2764 clonotypes. We identified multiple CDR3α, CDR3β, and paired CDR3α:CDR3β motif candidates. Among these, a previously known conserved CDR3β R-motif used by TRAV26-1/TRBV7-2 TCRs specific for the DQ2.5-glia-α2 epitope was the most prominent motif. Furthermore, we identified the epitope specificity of altogether 16 new CDR3α:CDR3β motifs by comparing with TCR sequences of 231 T-cell clones with known specificity and TCR sequences of cells sorted with single HLA-DQ2.5:gluten tetramers. We identified 325 public TCRα and TCRβ sequences of which 145, 102 and 78 belonged to TCRα, TCRβ and paired TCRαβ sequences, respectively. While the number of public sequences was depended on the number of clonotypes in each patient, we found that the proportion of public clonotypes from the gluten-specific TCR repertoire of given CeD patients appeared to be stable (median 37%). Taken together, we here demonstrate that the TCR repertoire of CD4+ T cells specific to immunodominant gluten epitopes in CeD is diverse, yet there is clearly biased V-gene usage, presence of public TCRs and existence of conserved motifs of which R-motif is the most prominent.

• Lemma, Roza Berhanu; Ledsaak, Marit; Fuglerud, Bettina Maria; Sandve, Geir Kjetil; Eskeland, Ragnhild & Gabrielsen, Odd Stokke (2021). Chromatin occupancy and target genes of the haematopoietic master transcription factor MYB. Scientific Reports. 11(9008). doi: 10.1038/s41598-021-88516-w. Fulltekst i vitenarkiv Vis sammendrag

  The transcription factor MYB is a master regulator in haematopoietic progenitor cells and a pioneer factor affecting differentiation and proliferation of these cells. Leukaemic transformation may be promoted by high MYB levels. Despite much accumulated molecular knowledge of MYB, we still lack a comprehensive understanding of its target genes and its chromatin action. In the present work, we performed a ChIP-seq analysis of MYB in K562 cells accompanied by detailed bioinformatics analyses. We found that MYB occupies both promoters and enhancers. Five clusters (C1–C5) were found when we classified MYB peaks according to epigenetic profiles. C1 was enriched for promoters and C2 dominated by enhancers. C2-linked genes were connected to hematopoietic specific functions and had GATA factor motifs as second in frequency. C1 had in addition to MYB-motifs a significant frequency of ETS-related motifs. Combining ChIP-seq data with RNA-seq data allowed us to identify direct MYB target genes. We also compared ChIP-seq data with digital genomic footprinting. MYB is occupying nearly a third of the super-enhancers in K562. Finally, we concluded that MYB cooperates with a subset of the other highly expressed TFs in this cell line, as expected for a master regulator

• Akbar, Rahmad; Robert, Philippe Paul Auguste; Pavlović, Milena; Jeliazkov, Jeliazko R.; Snapkov, Igor & Slabodkin, Andrei [Vis alle 15 forfattere av denne artikkelen] (2021). A compact vocabulary of paratope-epitope interactions enables predictability of antibody-antigen binding. Cell reports. 34:108856(11), s. 1–21. doi: 10.1016/j.celrep.2021.108856. Fulltekst i vitenarkiv Vis sammendrag

  Antibody-antigen binding relies on the specific interaction of amino acids at the paratope-epitope interface. The predictability of antibody-antigen binding is a prerequisite for de novo antibody and (neo-)epitope design. A fundamental premise for the predictability of antibody-antigen binding is the existence of paratope- epitope interaction motifs that are universally shared among antibody-antigen structures. In a dataset of non-redundant antibody-antigen structures, we identify structural interaction motifs, which together compose a commonly shared structure-based vocabulary of paratope-epitope interactions. We show that this vocabulary enables the machine learnability of antibody-antigen binding on the paratope-epitope level using generative machine learning. The vocabulary (1) is compact, less than 104 motifs; (2) distinct from non-immune protein-protein interactions; and (3) mediates specific oligo- and polyreactive interactions between paratope-epitope pairs. Our work leverages combined structure- and sequence-based learning to demonstrate that machine-learning-driven predictive paratope and epitope engineering is feasible.

• Greiff, Victor & Sandve, Geir Kjetil Ferkingstad (2021). Profiling the baseline performance and limits of machine learning models for adaptive immune receptor repertoire classification. BioRxiv. ISSN 0362-4331. Fulltekst i vitenarkiv
• Greiff, Victor & Sandve, Geir Kjetil Ferkingstad (2021). immuneML: an ecosystem for machine learning analysis of adaptive immune receptor repertoires. BioRxiv. ISSN 0362-4331. Fulltekst i vitenarkiv
• Greiff, Victor & Sandve, Geir Kjetil Ferkingstad (2021). One billion synthetic 3D-antibody-antigen complexes enable unconstrained machine-learning formalized investigation of antibody specificity prediction. BioRxiv. ISSN 0362-4331. doi: 10.1101/2021.07.06.451258. Fulltekst i vitenarkiv
• Khelik, Ksenia; Sandve, Geir Kjetil; Nederbragt, Alexander Johan & Rognes, Torbjørn (2020). NucBreak: location of structural errors in a genome assembly by using paired-end Illumina reads. BMC Bioinformatics. 21(1). doi: 10.1186/s12859-020-3414-0. Fulltekst i vitenarkiv Vis sammendrag

  Background Advances in whole genome sequencing strategies have provided the opportunity for genomic and comparative genomic analysis of a vast variety of organisms. The analysis results are highly dependent on the quality of the genome assemblies used. Assessment of the assembly accuracy may significantly increase the reliability of the analysis results and is therefore of great importance. Results Here, we present a new tool called NucBreak aimed at localizing structural errors in assemblies, including insertions, deletions, duplications, inversions, and different inter- and intra-chromosomal rearrangements. The approach taken by existing alternative tools is based on analysing reads that do not map properly to the assembly, for instance discordantly mapped reads, soft-clipped reads and singletons. NucBreak uses an entirely different and unique method to localise the errors. It is based on analysing the alignments of reads that are properly mapped to an assembly and exploit information about the alternative read alignments. It does not annotate detected errors. We have compared NucBreak with other existing assembly accuracy assessment tools, namely Pilon, REAPR, and FRCbam as well as with several structural variant detection tools, including BreakDancer, Lumpy, and Wham, by using both simulated and real datasets. Conclusions The benchmarking results have shown that NucBreak in general predicts assembly errors of different types and sizes with relatively high sensitivity and with lower false discovery rate than the other tools. Such a balance between sensitivity and false discovery rate makes NucBreak a good alternative to the existing assembly accuracy assessment tools and SV detection tools. NucBreak is freely available at https://github.com/uio-bmi/NucBreak under the MPL license.

• Grytten, Ivar; Rand, Knut Dagestad; Nederbragt, Alexander Johan & Sandve, Geir Kjetil (2020). Assessing graph-based read mappers against a baseline approach highlights strengths and weaknesses of current methods. BMC Genomics. 21. doi: 10.1186/s12864-020-6685-y. Fulltekst i vitenarkiv Vis sammendrag

  Background Graph-based reference genomes have become popular as they allow read mapping and follow-up analyses in settings where the exact haplotypes underlying a high-throughput sequencing experiment are not precisely known. Two recent papers show that mapping to graph-based reference genomes can improve accuracy as compared to methods using linear references. Both of these methods index the sequences for most paths up to a certain length in the graph in order to enable direct mapping of reads containing common variants. However, the combinatorial explosion of possible paths through nearby variants also leads to a huge search space and an increased chance of false positive alignments to highly variable regions. Results We here assess three prominent graph-based read mappers against a hybrid baseline approach that combines an initial path determination with a tuned linear read mapping method. We show, using a previously proposed benchmark, that this simple approach is able to improve overall accuracy of read-mapping to graph-based reference genomes. Conclusions Our method is implemented in a tool Two-step Graph Mapper, which is available at https://github.com/uio-bmi/two_step_graph_mapperalong with data and scripts for reproducing the experiments. Our method highlights characteristics of the current generation of graph-based read mappers and shows potential for improvement for future graph-based read mappers.

• Pré, Fleur du; Blazevski, Jana; Dewan, Alisa Elinsdatter; Stamnæs, Jorunn; Kanduri, Srinivasa Kalyana Chakravarthi & Sandve, Geir Kjetil [Vis alle 13 forfattere av denne artikkelen] (2020). B cell tolerance and antibody production to the celiac disease autoantigen transglutaminase 2. Journal of Experimental Medicine (JEM). ISSN 0022-1007. 217:e20190860.(2), s. 1–14. doi: 10.1084/jem.20190860. Fulltekst i vitenarkiv Vis sammendrag

  Autoantibodies to transglutaminase 2 (TG2) are hallmarks of celiac disease. To address B cell tolerance and autoantibody formation to TG2, we generated immunoglobulin knock-in (Ig KI) mice that express a prototypical celiac patient–derived anti-TG2 B cell receptor equally reactive to human and mouse TG2. We studied B cell development in the presence/absence of autoantigen by crossing the Ig KI mice to Tgm2−/− mice. Autoreactive B cells in Tgm2+/+ mice were indistinguishable from their naive counterparts in Tgm2−/− mice with no signs of clonal deletion, receptor editing, or B cell anergy. The autoreactive B cells appeared ignorant to their antigen, and they produced autoantibodies when provided T cell help. The findings lend credence to a model of celiac disease where gluten-reactive T cells provide help to autoreactive TG2-specific B cells by involvement of gluten–TG2 complexes, and they outline a general mechanism of autoimmunity with autoantibodies being produced by ignorant B cells on provision of T cell help.

• Weber, Cédric R.; Akbar, Rahmad; Yermanos, Alexander; Pavlovic, Milena; Snapkov, Igor & Sandve, Geir Kjetil [Vis alle 8 forfattere av denne artikkelen] (2020). immuneSIM: tunable multi-feature simulation of B- and T-cell receptor repertoires for immunoinformatics benchmarking. Bioinformatics. ISSN 1367-4803. 36(11), s. 3594–3596. doi: 10.1093/bioinformatics/btaa158. Fulltekst i vitenarkiv Vis sammendrag

  Abstract Summary B- and T-cell receptor repertoires of the adaptive immune system have become a key target for diagnostics and therapeutics research. Consequently, there is a rapidly growing number of bioinformatics tools for immune repertoire analysis. Benchmarking of such tools is crucial for ensuring reproducible and generalizable computational analyses. Currently, however, it remains challenging to create standardized ground truth immune receptor repertoires for immunoinformatics tool benchmarking. Therefore, we developed immuneSIM, an R package that allows the simulation of native-like and aberrant synthetic full-length variable region immune receptor sequences by tuning the following immune receptor features: (i) species and chain type (BCR, TCR, single and paired), (ii) germline gene usage, (iii) occurrence of insertions and deletions, (iv) clonal abundance, (v) somatic hypermutation and (vi) sequence motifs. Each simulated sequence is annotated by the complete set of simulation events that contributed to its in silico generation. immuneSIM permits the benchmarking of key computational tools for immune receptor analysis, such as germline gene annotation, diversity and overlap estimation, sequence similarity, network architecture, clustering analysis and machine learning methods for motif detection. Availability and implementation The package is available via https://github.com/GreiffLab/immuneSIM and on CRAN at https://cran.r-project.org/web/packages/immuneSIM. The documentation is hosted at https://immuneSIM.readthedocs.io. Contact sai.reddy@ethz.ch or victor.greiff@medisin.uio.no Supplementary information Supplementary data are available at Bioinformatics online.

• Salvatore, Stefania; Rand, Knut Dagestad; Grytten, Ivar; Ferkingstad, Egil; Domanska, Diana & Holden, Lars [Vis alle 10 forfattere av denne artikkelen] (2019). Beware the Jaccard: the choice of similarity measure is important and non-trivial in genomic colocalisation analysis. Briefings in Bioinformatics. ISSN 1467-5463. s. 1–8. doi: 10.1093/bib/bbz083. Fulltekst i vitenarkiv
• Brown, Alex; Snapkov, Igor; Akbar, Rahmad; Pavlovic, Milena; Miho, Enkelejda & Sandve, Geir Kjetil [Vis alle 7 forfattere av denne artikkelen] (2019). Augmenting adaptive immunity: progress and challenges in the quantitative engineering and analysis of adaptive immune receptor repertoires. Molecular Systems Design & Engineering. 4(4), s. 701–736. doi: 10.1039/c9me00071b. Fulltekst i vitenarkiv
• Kanduri, Srinivasa Kalyana Chakravarthi; Bock, Christoph; Gundersen, Sveinung; Hovig, Eivind & Sandve, Geir Kjetil (2019). Colocalization analyses of genomic elements: approaches, recommendations and challenges. Bioinformatics. ISSN 1367-4803. 35(9), s. 1615–1624. doi: 10.1093/bioinformatics/bty835. Fulltekst i vitenarkiv Vis sammendrag

  Motivation: Many high-throughput methods produce sets of genomic regions as one of their main outputs. Scientists often use genomic colocalization analysis to interpret such region sets, for example to identify interesting enrichments and to understand the interplay between the underlying biological processes. Although widely used, there is little standardization in how these analyses are performed. Different practices can substantially affect the conclusions of colocalization analyses. Results: Here, we describe the different approaches and provide recommendations for performing genomic colocalization analysis, while also discussing common methodological challenges that may influence the conclusions. As illustrated by concrete example cases, careful attention to analysis details is needed in order to meet these challenges and to obtain a robust and biologically meaningful interpretation of genomic region set data.

• Grytten, Ivar; Rand, Knut Dagestad; Nederbragt, Alexander Johan; Storvik, Geir Olve; Glad, Ingrid Kristine & Sandve, Geir Kjetil (2019). Graph Peak Caller: Calling ChIP-seq peaks on graph-based reference genomes. PLoS Computational Biology. ISSN 1553-734X. 15(2), s. 1–13. doi: 10.1371/journal.pcbi.1006731. Fulltekst i vitenarkiv Vis sammendrag

  Graph-based representations are considered to be the future for reference genomes, as they allow integrated representation of the steadily increasing data on individual variation. Currently available tools allow de novo assembly of graph-based reference genomes, alignment of new read sets to the graph representation as well as certain analyses like variant calling and haplotyping. We here present a first method for calling ChIP-Seq peaks on read data aligned to a graph-based reference genome. The method is a graph generalization of the peak caller MACS2, and is implemented in an open source tool, Graph Peak Caller. By using the existing tool vg to build a pan-genome of Arabidopsis thaliana, we validate our approach by showing that Graph Peak Caller with a pan-genome reference graph can trace variants within peaks that are not part of the linear reference genome, and find peaks that in general are more motif-enriched than those found by MACS2.

• Snir, Omri; Kanduri, Srinivasa Kalyana Chakravarthi; Lundin, Knut Erik Aslaksen; Sandve, Geir Kjetil & Sollid, Ludvig Magne (2019). Transcriptional profiling of human intestinal plasma cells reveals effector functions beyond antibody production. United European Gastroenterology Journal (UEG Journal). ISSN 2050-6406. 0(0), s. 1–9. doi: 10.1177/2050640619862461. Fulltekst i vitenarkiv
• Gheorghe, Marius; Sandve, Geir Kjetil; Khan, Aziz; Cheneby, Jeanne; Ballester, Benoit & Mathelier, Anthony (2019). A map of direct TF–DNA interactions in the human genome. Nucleic Acids Research (NAR). ISSN 0305-1048. 47(4), s. e21–e21. doi: 10.1093/nar/gky1210. Fulltekst i vitenarkiv Vis sammendrag

  Chromatin immunoprecipitation followed by sequencing (ChIP-seq) is the most popular assay to identify genomic regions, called ChIP-seq peaks, that are bound in vivo by transcription factors (TFs). These regions are derived from direct TF–DNA interactions, indirect binding of the TF to the DNA (through a co-binding partner), nonspecific binding to the DNA, and noise/bias/artifacts. Delineating the bona fide direct TF–DNA interactions within the ChIP-seq peaks remains challenging. We developed a dedicated software, ChIP-eat, that combines computational TF binding models and ChIP-seq peaks to automatically predict direct TF–DNA interactions. Our work culminated with predicted interactions covering >2% of the human genome, obtained by uniformly processing 1983 ChIP-seq peak data sets from the ReMap database for 232 unique TFs. The predictions were a posteriori assessed using protein binding microarray and ChIP-exo data, and were predominantly found in high quality ChIP-seq peaks. The set of predicted direct TF–DNA interactions suggested that high-occupancy target regions are likely not derived from direct binding of the TFs to the DNA. Our predictions derived co-binding TFs supported by protein-protein interaction data and defined cis-regulatory modules enriched for disease- and trait-associated SNPs. We provide this collection of direct TF–DNA interactions and cis-regulatory modules through the UniBind web-interface (http://unibind.uio.no).

• Simovski, Boris; Kanduri, Srinivasa Kalyana Chakravarthi; Gundersen, Sveinung; Titov, Dmytro; Domanska, Diana Ewa & Bock, Christoph [Vis alle 15 forfattere av denne artikkelen] (2018). Coloc-stats: A unified web interface to perform colocalization analysis of genomic features. Nucleic Acids Research (NAR). ISSN 0305-1048. 46(1), s. W186–W193. doi: 10.1093/nar/gky474. Fulltekst i vitenarkiv Vis sammendrag

  Functional genomics assays produce sets of genomic regions as one of their main outputs. To biologically interpret such region-sets, researchers often use colocalization analysis, where the statistical significance of colocalization (overlap, spatial proximity) between two or more region-sets is tested. Existing colocalization analysis tools vary in the statistical methodology and analysis approaches, thus potentially providing different conclusions for the same research question. As the findings of colocalization analysis are often the basis for follow-up experiments, it is helpful to use several tools in parallel and to compare the results. We developed the Coloc-stats web service to facilitate such analyses. Coloc-stats provides a unified interface to perform colocalization analysis across various analytical methods and method-specific options (e.g. colocalization measures, resolution, null models). Coloc-stats helps the user to find a method that supports their experimental requirements and allows for a straightforward comparison across methods. Coloc-stats is implemented as a web server with a graphical user interface that assists users with configuring their colocalization analyses.

• Risnes, Louise Fremgaard; Christophersen, Asbjørn; Dahal-Koirala, Shiva; Neumann, Ralf Stefan; Sandve, Geir Kjetil & Sarna, Vikas Kumar [Vis alle 9 forfattere av denne artikkelen] (2018). Disease-driving CD4+ T cell clonotypes persist for decades in celiac disease. Journal of Clinical Investigation. ISSN 0021-9738. 128(6), s. 2642–2650. doi: 10.1172/JCI98819. Fulltekst i vitenarkiv Vis sammendrag

  Little is known about the repertoire dynamics and persistence of pathogenic T cells in HLA-associated disorders. In celiac disease, a disorder with a strong association with certain HLA-DQ allotypes, presumed pathogenic T cells can be visualized and isolated with HLA-DQ:gluten tetramers, thereby enabling further characterization. Single and bulk populations of HLA-DQ:gluten tetramer–sorted CD4+ T cells were analyzed by high-throughput DNA sequencing of rearranged TCR-α and -β genes. Blood and gut biopsy samples from 21 celiac disease patients, taken at various stages of disease and in intervals of weeks to decades apart, were examined. Persistence of the same clonotypes was seen in both compartments over decades, with up to 53% overlap between samples obtained 16 to 28 years apart. Further, we observed that the recall response following oral gluten challenge was dominated by preexisting CD4+ T cell clonotypes. Public features were frequent among gluten-specific T cells, as 10% of TCR-α, TCR-β, or paired TCR-αβ amino acid sequences of total 1813 TCRs generated from 17 patients were observed in 2 or more patients. In established celiac disease, the T cell clonotypes that recognize gluten are persistent for decades, making up fixed repertoires that prevalently exhibit public features. These T cells represent an attractive therapeutic target.

• Domanska, Diana Ewa; Kanduri, Srinivasa Kalyana Chakravarthi; Simovski, Boris & Sandve, Geir Kjetil (2018). Mind the gaps: overlooking inaccessible regions confounds statistical testing in genome analysis. BMC Bioinformatics. 19(481). doi: 10.1186/s12859-018-2438-1. Fulltekst i vitenarkiv Vis sammendrag

  Background The current versions of reference genome assemblies still contain gaps represented by stretches of Ns. Since high throughput sequencing reads cannot be mapped to those gap regions, the regions are depleted of experimental data. Moreover, several technology platforms assay a targeted portion of the genomic sequence, meaning that regions from the unassayed portion of the genomic sequence cannot be detected in those experiments. We here refer to all such regions as inaccessible regions, and hypothesize that ignoring these regions in the null model may increase false findings in statistical testing of colocalization of genomic features. Results Our explorative analyses confirm that the genomic regions in public genomic tracks intersect very little with assembly gaps of human reference genomes (hg19 and hg38). The little intersection was observed only at the beginning and end portions of the gap regions. Further, we simulated a set of synthetic tracks by matching the properties of real genomic tracks in a way that nullified any true association between them. This allowed us to test our hypothesis that not avoiding inaccessible regions (as represented by assembly gaps) in the null model would result in spurious inflation of statistical significance. We contrasted the distributions of test statistics and p-values of Monte Carlo-based permutation tests that either avoided or did not avoid assembly gaps in the null model when testing colocalization between a pair of tracks. We observed that the statistical tests that did not account for assembly gaps in the null model resulted in a distribution of the test statistic that is shifted to the right and a distribution of p-values that is shifted to the left (indicating inflated significance). We observed a similar level of inflated significance in hg19 and hg38, despite assembly gaps covering a smaller proportion of the latter reference genome. Conclusion We provide empirical evidence demonstrating that inaccessible regions, even when covering only a few percentages of the genome, can lead to a substantial amount of false findings if not accounted for in statistical colocalization analysis.

• Yao, Ying; Zia, Asima; Wyrozemski, Lukasz Adam; Lindeman, Ida; Sandve, Geir Kjetil & Qiao, Shuo Wang (2018). Exploiting antigen receptor information to quantify index switching in single-cell transcriptome sequencing experiments. PLOS ONE. 13(12), s. 1–17. doi: 10.1371/journal.pone.0208484. Fulltekst i vitenarkiv Vis sammendrag

  By offering high sequencing speed and ultra-high-throughput at a low price, Illumina next-generation sequencing platforms have been widely adopted in recent years. However, an experiment with multiplexed library could be at risk of molecular recombination, known as “index switching”, which causes a proportion of the reads to be assigned to an incorrect sample. It is reported that a new advance, exclusion amplification (ExAmp) in conjunction with the patterned flow cell technology introduced on HiSeq 3000/HiSeq 4000/HiSeq X sequencing systems, potentially suffers from a higher rate of index switching than conventional bridge amplification. We took advantage of the diverse but highly cell-specific expression of antigen receptors on immune cells to quantify index switching on single cell RNA-seq data that were sequenced on HiSeq 3000 and HiSeq 4000. By utilizing the unique antigen receptor expression, we could quantify the spread-of-signal from many different wells (n = 55 from total of three batches) due to index switching. Based on full-length T cell receptor (TCR) sequences from all samples reconstructed by TraCeR and TCR gene expression quantified by Kallisto, we found index switching in all three batches of experiments investigated. The median percentage of incorrectly detected markers was estimated to be 3.9% (interquartile range (IQR): 1.7%-7.3%). We did not detect any consistent patterns of certain indices to be more prone to switching than others, suggesting that index switching is a stochastic process. Our results confirm that index switching is a problem that affects samples run in multiplexed libraries on Illumina HiSeq 3000 and HiSeq 4000 platforms.

Se alle arbeider i NVA

• Pollock, Joanna Katharine Mary & Sandve, Geir Kjetil Ferkingstad (2026). Diskusjon om kunstig intelligens i forskning. Fulltekst i vitenarkiv
• Jensenius, Alexander Refsum; Sandve, Geir Kjetil Ferkingstad; Boasson, Elin Lerum; Mjaaland, Marius Timmann; Skjelkvåle, Brit Lisa & Moen, Kåre [Vis alle 7 forfattere av denne artikkelen] (2026). What can the Centre for Global Sustainability do for UiO – and the other way around? Fulltekst i vitenarkiv
• Malthe-Sørenssen, Anders; Harang, Fabian Andsem; Johnson, Victoria Christine Årang; Nyrup, Jacob; Lie, Knut-Andreas & Sandve, Geir Kjetil Ferkingstad [Vis alle 7 forfattere av denne artikkelen] (2026). Hvordan endrer KI-agenter forskningsprosessen? Fulltekst i vitenarkiv Vis sammendrag

  Dette møtet vil ta for seg de forskningsetiske spørsmålene som reiser seg i møte med bruken av KI-agenter i forskning

• Jemterud, Torkild; Roberts, Aubrey Jane; Sandve, Geir Kjetil Ferkingstad & Vestre, Katharina (2026). Hva er greia med hestesko? - Abels tårn - NRK Radio. [Radio]. NRK. Fulltekst i vitenarkiv Vis sammendrag

  Når vet en stamcelle hva den skal bli? Hvordan belønne KI? Er jeg mer av min mormor enn av min mor? Hadde dinosaurene pandemier? Hvor fargeblinde er hunder?

• Skogvoll, Vidar; Sandve, Geir Kjetil Ferkingstad; Kempton, Alexander Moltubakk & Gregers, Tone Fredsvik (2024). Real læring - Realistfabrikken: Hvordan lage informatiker. [Radio]. Podcast. Fulltekst i vitenarkiv
• Nickelsen, Trine & Sandve, Geir Kjetil Ferkingstad (2024). Kunstig intelligens: Avslører sykdom før symptomene kommer. [Journal]. Apollon. Fulltekst i vitenarkiv
• Rand, Knut Dagestad; Grytten, Ivar; Pavlović, Milena; Kanduri, Chakravarthi & Sandve, Geir Kjetil Ferkingstad (2024). BioNumPy: array programming for biology. Nature Methods. ISSN 1548-7091. doi: 10.1038/s41592-024-02483-4. Fulltekst i vitenarkiv
• Torgersen, Eivind; Rand, Knut Dagestad; Grytten, Ivar & Sandve, Geir Kjetil Ferkingstad (2024). Nå blir det enklere å kode for forskere som jobber med biologiske data. [Journal]. Digi.no. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2023). Deciphering how adaptive immune cells recognise pathogens: gathering suited data, defining appropriate assessments and incrementally improving machine learning methodology. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2023). Hvorfor er maskinlæring viktig i medisin? Fulltekst i vitenarkiv
• Vu, Mai Ha; Akbar, Rahmad; Robert, Philippe; Sandve, Geir Kjetil Ferkingstad; Haug, Dag Trygve Truslew & Greiff, Victor (2023). Linguistically inspired roadmap for building biologically reliable protein language models. Nature Machine Intelligence. 5(5), s. 485–496. doi: 10.1038/s42256-023-00637-1. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2023). Approaching intriguing problems with machine learning: the full picture from data availability to methodology development and assessment, with the adaptive immune system as case. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad & Sprenger, Kayla (2023). Computational Era of Life Sciences. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2023). Collectively deciphering the rules of immune receptor-antigen binding: deeply analyse the problem, obtain the necessary data, define standardised benchmarks, and ensure effective method comparison. Fulltekst i vitenarkiv
• Dæhlen, Morten; Sandve, Geir Kjetil Ferkingstad & Braa, Kristin (2023). KI-milliarden må brukes på en rettferdig grønn og digital omstilling. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad & Pensar, Johan (2022). Machine Learning and Causality. Fulltekst i vitenarkiv
• Ferenc, Katalin Terezia; Hsieh, Ping-Han; Mathelier, Anthony & Sandve, Geir Kjetil Ferkingstad (2022). Method development for the identification of cis-regulatory signatures of cell types. Fulltekst i vitenarkiv
• Ferenc, Katalin Terezia; Mathelier, Anthony; Hsieh, Ping-Han & Sandve, Geir Kjetil Ferkingstad (2022). Method development for the identification of cis-regulatory signatures of cell types. Fulltekst i vitenarkiv
• Ferenc, Katalin Terezia; Hsieh, Ping-Han; Mathelier, Anthony & Sandve, Geir Kjetil Ferkingstad (2022). Method development for the identification of cis-regulatory signatures of cell types. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad; Greiff, Victor; Haff, Ingrid Hobæk & Vu, Mai Ha (2022). Deciphering the Immune System Through Linguistics-Inspired Statistical Machine Learning. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2022). Det inspirerende og frustrerende ved at programmering er en kreativ prosess. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2022). Are quick and dirty programming habits sufficient in science - and do we at all need to consider code quality in science education? Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2022). Kunstig intelligens viser oss hva immunforsvaret jobber med. Fulltekst i vitenarkiv
• Dance, Amber; Sandve, Geir Kjetil Ferkingstad; Morris, Samantha; Saez-Rodriguez, Julio; Heyn, Holger & Levin, Joshua (2022). Which single-cell analysis tool is best? Scientists offer advice. [Journal]. Nature. doi: 10.1038/d41586-022-04426-5. Fulltekst i vitenarkiv
• Doeland, Elin Martine; Sandve, Geir Kjetil Ferkingstad & Greiff, Victor (2022). Immunforsvaret har skjulte mønstre om sykdom og infeksjoner. Forskning.no. ISSN 1891-635X. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2022). How the choice of programming language matters when learning programming - Python, Java, R and more. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad & Greiff, Victor (2022). Access to ground truth at unconstrained size makes simulated data as indispensable as experimental data for bioinformatics methods development and benchmarking. Bioinformatics. ISSN 1367-4803. 38(21), s. 4994–4996. doi: 10.1093/bioinformatics/btac612. Fulltekst i vitenarkiv
• Balaban, Gabriel; Grytten, Ivar; Rand, Knut Dagestad; Scheffer, Lonneke & Sandve, Geir Kjetil (2021). Ten simple rules for quick and dirty scientific programming. PLoS Computational Biology. ISSN 1553-734X. 17:e1008549(3), s. 1–15. doi: 10.1371/journal.pcbi.1008549. Fulltekst i vitenarkiv
• Sandve, Geir Kjetil Ferkingstad (2021). ImmuneML and the case for domain-tailored machine learning platforms. Fulltekst i vitenarkiv
• Kanduri, Chakravarthi; Sandve, Geir Kjetil Ferkingstad; Hovig, Eivind; De, Subhajyoti & Layer, Ryan (2021). Editorial: Genomic Colocalization and Enrichment Analyses. Frontiers in Genetics. 11. doi: 10.3389/fgene.2020.617876. Fulltekst i vitenarkiv
• Widrich, Michael; Schäfl, Bernhard; Pavlović, Milena; Ramsauer, Hubert; Gruber, Lukas & Holzleitner, Markus [Vis alle 11 forfattere av denne artikkelen] (2020). Modern Hopfield Networks and Attention for Immune Repertoire Classification. Fulltekst i vitenarkiv Vis sammendrag

  A central mechanism in machine learning is to identify, store, and recognize patterns. How to learn, access, and retrieve such patterns is crucial in Hopfield networks and the more recent transformer architectures. We show that the attention mechanism of transformer architectures is actually the update rule of modern Hopfield networks that can store exponentially many patterns. We exploit this high storage capacity of modern Hopfield networks to solve a challenging multiple instance learning (MIL) problem in computational biology: immune repertoire classification. In immune repertoire classification, a vast number of immune receptors are used to predict the immune status of an individual. This constitutes a MIL problem with an unprecedentedly massive number of instances, two orders of magnitude larger than currently considered problems, and with an extremely low witness rate. Accurate and interpretable machine learning methods solving this problem could pave the way towards new vaccines and therapies, which is currently a very relevant research topic intensified by the COVID-19 crisis. In this work, we present our novel method DeepRC that integrates transformer-like attention, or equivalently modern Hopfield networks, into deep learning architectures for massive MIL such as immune repertoire classification. We demonstrate that DeepRC outperforms all other methods with respect to predictive performance on large-scale experiments including simulated and real-world virus infection data and enables the extraction of sequence motifs that are connected to a given disease class. Source code and datasets: https://github.com/ml-jku/DeepRC.

• Molyneux, Sam D.; Waterhouse, Paul D.; Shelton, Dawne; Shao, Yang W.; Watling, Christopher M. & Tang, Qing-Lian [Vis alle 18 forfattere av denne artikkelen] (2020). Author Correction: Human somatic cell mutagenesis creates genetically tractable sarcomas (Nature Genetics, (2014), 46, 9, (964-972), 10.1038/ng.3065). Nature Genetics. ISSN 1061-4036. 52(4), s. 464–464. doi: 10.1038/s41588-020-0589-2. Fulltekst i vitenarkiv
• Gheorghe, Marius; Sandve, Geir Kjetil Ferkingstad; Khan, Aziz; Chèneby, Jeanne; Ballester, Benoit & Mathelier, Anthony (2019). Erratum: A map of direct TF-DNA interactions in the human genome (Nucleic acids research (2019) 47 4 (e21)). Nucleic Acids Research (NAR). ISSN 0305-1048. 47(14), s. 7715–7715. doi: 10.1093/nar/gkz582. Fulltekst i vitenarkiv
• Kanduri, Chakravarthi; Domanska, Diana; Hovig, Eivind & Sandve, Geir Kjetil (2017). Genome build information is an essential part of genomic track files. Genome Biology. ISSN 1465-6906. 18(1). doi: 10.1186/s13059-017-1312-1. Fulltekst i vitenarkiv Vis sammendrag

  Genomic locations are represented as coordinates on a specific genome build version, but the build information is frequently missing when coordinates are provided. We show that this information is essential to correctly interpret and analyse the genomic intervals contained in genomic track files. Although not a substitute for best practices, we also provide a tool to predict the genome build version of genomic track files.

• Shivashankar, Karthik; Martini, Antonio & Sandve, Geir Kjetil Ferkingstad (2025). The Dual Role of Machine Learning in Technical Debt Management. Applying Machine Learning for Identification While Analyzing Debt within ML Systems. Universitetet i Oslo. Fulltekst i vitenarkiv Vis sammendrag

  Context: The management of Technical Debt (TD) is a critical challenge in software engineering, directly impacting project costs and long-term viability. The proliferation of Machine Learning (ML) systems introduces a complex duality: they present novel forms of debt while simultaneously offering powerful automated techniques for its identification and mitigation. Problem: Existing TD management practices often rely on manual, subjective processes or generic tools that fail to scale or address the specific quality challenges inherent in modern codebases, particularly those involving Python and ML. A comprehensive, automated, and context-aware approach is needed to effectively navigate this dual role of ML in TD. Methodology: This thesis addresses this gap through a multi-faceted research program. We developed a suite of automated tools leveraging Advanced Natural Language Processing (NLP) with transformer models for analyzing textual artifacts (e.g., issue reports) and domain-specific static code analysis for dissecting Python and ML codebases. This practical tool development is complemented by a systematic literature review to map the landscape of quality challenges in ML systems and an exploratory study on using Large Language Models (LLMs) for proactive code improvement. Contributions: The research yields six interconnected contributions. We demonstrate that specialized, fine-tuned transformer models (BEAConTD, QualiTagger) consistently outperform larger, general-purpose models for classifying TD and ISO 25010 quality attributes. We introduce domain-specific static analysis tools, PyExamine and MLScent, that effectively detect 49 traditional Python smells and 76 unique ML-specific anti-patterns, respectively. Further, we establish that fine-tuned LLMs can verifiably improve code maintainability metrics. These practical tools are underpinned by a comprehensive mapping of 41 maintainability and 13 scalability challenges in ML workflows, providing a foundational reference for the field. All tools and datasets are released as open-source artifacts. Implications: Collectively, this work represents a significant advancement in automated software quality management. It provides practitioners with a suite of validated tools (TD-Suite, PyExamine, MLScent, QualiTagger) for objective, scalable, and data-driven TD management. The findings establish new research directions at the intersection of software engineering and artificial intelligence, enabling quality control that can keep pace with the complexity of modern software.

Se alle arbeider i NVA