Abstract
Combinatorial treatments involving two or more drugs have become a standard of care for various complex diseases, including tuberculosis, malaria, HIV and other viral infections, as well as most of the advanced cancers. High-throughput screening in preclinical model systems (e.g. cancer cell lines or viral infection models) is the state-of-the-art approach to systematically identify candidate drug combinations. However, due to the exponential number of possible drug combinations and the extensive heterogeneity of the target systems, computational methods, in particular machine learning, are critically needed to guide the discovery of effective combinations to be prioritized for further pre-clinical validation and clinical development.
In this talk, I will present comboFM [1], a novel machine learning framework for predicting the responses of drug combinations in preclinical studies. comboFM models the cell context-specific drug interactions through higher-order tensors, and efficiently learns latent factors of the tensor using powerful factorization machines. The approach enables comboFM to leverage information from previous experiments performed on similar drugs and cells when predicting responses of new combinations in so far untested cells; thereby, it achieves highly accurate predictions despite sparsely populated data tensors.
High predictive performance of comboFM is demonstrated in various prediction scenarios using data from cancer cell line drug screening. Subsequent experimental validation of a set of previously untested drug combinations further supports the practical and robust applicability of comboFM.
[1] Julkunen, H.J., Cichonska, A., Gautam, P., Szedmak, S., Douat, J., Pahikkala, T., Aittokallio, T. and Rousu, J., 2020. comboFM: leveraging multi-way interactions for systematic prediction of drug combination effects. bioRxiv https://doi.org/10.1101/2020.09.02.278986
Zoom connection information
Junior talk
Erica Ponzi, postdoctoral fellow at OCBE, will present her work on "Integrative analysis of multi-omics data improves model predictions: an application to lung cancer."