Anticancer drug synergy prediction in understudied tissues using transfer learning

Objective Drug combination screening has advantages in identifying cancer treatment options with higher efficacy without degradation in terms of safety. A key challenge is that the accumulated number of observations in in-vitro drug responses varies greatly among different cancer types, where some tissues are more understudied than the others. Thus, we aim to develop a drug synergy prediction model for understudied tissues as a way of overcoming data scarcity problems. Materials and Methods We collected a comprehensive set of genetic, molecular, phenotypic features for cancer cell lines. We developed a drug synergy prediction model based on multitask deep neural networks to integrate multimodal input and multiple output. We also utilized transfer learning from data-rich tissues to data-poor tissues. Results We showed improved accuracy in predicting synergy in both data-rich tissues and understudied tissues. In data-rich tissue, the prediction model accuracy was 0.9577 AUROC for binarized classification task and 174.3 mean squared error for regression task. We observed that an adequate transfer learning strategy significantly increases accuracy in the understudied tissues. Conclusions Our synergy prediction model can be used to rank synergistic drug combinations in understudied tissues and thus help to prioritize future in-vitro experiments. Code is available at https://github.com/yejinjkim/synergy-transfer.

Anti-cancer Drug Synergy Prediction in Understudied Tissues using Transfer Learning

MotivationExploring an exponentially increasing yet more promising space, high-throughput combinatorial drug screening has advantages in identifying cancer treatment options with higher efficacy without degradation in terms of safety. A key challenge is that accumulated number of observations in in-vitro drug responses varies greatly among different cancer types, where some tissues (such as bone and prostate) are understudied than the others. Thus, we aim to develop a drug synergy prediction model for understudied data-poor tissues as overcoming data scarcity problem.ResultsWe collected a comprehensive set of genetic, molecular, phenotypic features for cancer cell lines from six different databases. We developed a drug synergy prediction model based on deep neural networks to integrate multi-modal input and utilize transfer learning from data-rich tissues to data-poor tissues. We showed improved accuracy in predicting drug synergy in understudied tissues without enough drug combination screening data nor after-treatment transcriptome. Our synergy prediction model can be used to rank synergistic drug combinations in understudied tissues and thus help prioritizing future in-vitro experiments.Availability and ImplementationOur algorithm will be publicly available via https://github.com/yejinjkim/drug-synergy-prediction

Stratification and prediction of drug synergy based on target functional similarity

ABSTRACTTargeted mono-therapies in cancer are hampered by the ability of tumor cells to escape inhibition through rewiring or alternative pathways. Drug combination approaches can provide a means to overcome these resistance mechanisms. Effective use of combinations requires strategies to select combinations from the enormous space of combinations, and to stratify patients according to their likelihood to respond. We here introduce two complementary workflows: One prioritising experiments in high-throughput screens for drug synergy enrichment, and a consecutive workflow to predict hypothesis-driven synergy stratification. Both approaches only need data of efficacy of single drugs. They rely on the notion of target functional similarity between two target proteins. This notion reflects how similarly effective drugs are on different cancer cells as a function of cancer signaling pathways’ activities on those cells. Our synergy prediction workflow revealed that two drugs targeting either the same or functionally opposite pathways are more likely to be synergistic. This enables experimental prioritisation in high-throughput screens and supports the notion that synergy can be achieved by either redundant pathway inhibition or targeting independent compensatory mechanisms. We tested the synergy stratification workflow on seven target protein pairs (AKT/EGFR, AKT/MTOR, BCL2/MTOR, EGFR/MTOR, AKT/BCL2, AKT/ALK and AKT/PARP1, representing 29 combinations and predicted their synergies in 33 breast cancer cell lines (Pearson’s correlation r=0.27). Additionally, we experimentally validated predicted synergy of the BRAF/Insulin Receptor combination (Dabrafenib/BMS−754807) in 48 colorectal cancer cell lines (r=0.5). In conclusion, our synergy prediction workflow can support compound prioritization in large scale drug screenings, and our synergy stratification workflow can select where the efficacy of drugs already known for inducing synergy is higher.