Predicting drug response using biological networks

Abstract

It is necessary to predict the efficacy of individual drugs on patients to realize personalized medicine. Testing drugs on patients in clinical trial is the only way to evaluate the efficacy of drugs. The approach is labor in-tensive and requires overwhelming costs and a number of experiments. Therefore, preclinical model system has been intensively investigated for predicting the efficacy of drugs. To predict the efficacy of drugs, we de-veloped two different methods that are simulation-based method and machine learning-based method. In simulation-based method, we made Multi-Level Modeling Language (MLML) for modeling biological systems, which is multi-scale systems. MLML contain context information that is spatial scale, temporal scale and condition information, such as disease. We have applied MLML to model type 2 diabetes (T2D), which involves the malfunction of numerous organs such as pancreas, liver, and muscle. We have extracted automatically T2D-related 12,522 rules from public DB. We simulated response of single drugs and combina-tion drugs on T2D model by petri-net simulation. The results of our simulation show T2D candidate drugs and how combination drugs could work on the whole-body scale In machine learning-based method, we developed cell line specific functional modules, which are cell line specific features, as prediction features to predict the efficacy of drugs. Cell line specific functional modules are clusters of genes, which have similar biological functions in cell line specific networks. We used two types of data from the US National Cancer Institute 60 anticancer drug screen (NCI60)) cell line transcription data of nine tissue types and 52 different cell lines, 2) drug response data of 30 drugs across 52 of the cell lines. We used linear regression for drug sensitivity prediction. We assessed the prediction performance in leave-one-out cross-validation (LOOCV). We also selected functions which are associated with drug sensitivity. Our method performed better than gene-based model. We also analyzed selected functions, which are asso-ciated with drug sensitivity, of five drugs - lapatinib, erotinib, raloxifen, tamoxifen and gefitinib- by our mod-el. Two drug pairs in five drugs have same therapeutic effect. Our model also showed that two drug pairs have same selected functions. These results suggest that our model can provide drug sensitivity prediction and also provide functions which are associated with drug sensitivity. Therefore, we could utilize drug sensitivity associated functions for drug repositioning or for suggesting secondary drugs for overcoming drug resistance.