Predicting Anticancer Drug Responses Using a Dual-Layer Integrated Cell Line-Drug Network Model

Abstract

The ability to predict the response of a cancer patient to a therapeutic agent is a major goal in modern oncology that should ultimately lead to personalized treatment. Existing approaches to predicting drug sensitivity rely primarily on profiling of cancer cell line panels that have been treated with different drugs and selecting genomic or functional genomic features to regress or classify the drug response. Here, we propose a dual-layer integrated cell line-drug network model, which uses both cell line similarity network (CSN) data and drug similarity network (DSN) data to predict the drug response of a given cell line using a weighted model. Using the Cancer Cell Line Encyclopedia (CCLE) and Cancer Genome Project (CGP) studies as benchmark datasets, our single-layer model with CSN or DSN and only a single parameter achieved a prediction performance comparable to the previously generated elastic net model. When using the dual-layer model integrating both CSN and DSN, our predicted response reached a 0.6 Pearson correlation coefficient with observed responses for most drugs, which is significantly better than the previous results using the elastic net model. We have also applied the dual-layer cell line-drug integrated network model to fill in the missing drug response values in the CGP dataset. Even though the dual-layer integrated cell line-drug network model does not specifically model mutation information, it correctly predicted that BRAF mutant cell lines would be more sensitive than BRAF wild-type cell lines to three MEK1/2 inhibitors tested.

Drug Response Prediction as a Link Prediction Problem

“…Zhang et al . constructed a dual-layered network consisting of cell lines and drugs where weighted edges between two nodes of the same type represented similarity and edges between cell lines and drugs contained the response data10. While the application of network-based approaches to drug response prediction is limited, network-based methods, specifically those employing link prediction, have been successful in many biomedical problems that involve predictive tasks.…”

Multitask learning improves prediction of cancer drug sensitivity

“…Two existing methods have reported outperforming elastic net, Bayesian multitask multiple kernel learning13 and dual-layer integrated cell line-drug neural network model27. While both methods demonstrated a performance advantage over elastic net, they were both kernel based, and therefore these prediction models have limited interpretability.…”

Inferences of individual drug responses across diverse cancer types using a novel competing endogenous RNA network

“…It was hypothesized that cancer patients with similar expression patterns would respond similarly to drugs having similar chemical structures (Zhang et al ., 2015). Using the patient–drug two‐layer integrated network, a linear weighting method was developed to predict the DRS of patient P t to drug D k based on an actual patient–drug relationship ( P i , D j ) obtained from clinical treatment records.$\text{DRS}false(P_t,D_{k}false)=\frac{\sum _j\sum _{i}f({\mathrm{normal\omega }}_{P_t{P}_i})\ast \text{DRS}(P_i,D_j)\ast f({\mathrm{\omega }}_{D_j{D}_k}false)}{\underset{j}{false\sum }\underset{i}{false\sum }ffalse({\mathrm{\omega }}_{P_t{P}_i}false)\ast ffalse({\mathrm{\omega }}_{D_j{D}_k}false)},$where DRS ( P t , D k ) was the DRS of patient P t to drug D k , ${\mathrm{\omega }}_{P_t{P}_i}$ and ${\mathrm{\omega }}_{D_J{D}_k}$ were the edge weights of P t − P i and D j − D k in the network which were converted by function $ffalse({\mathrm{\omega }}_{P_t{P}_i}false)=e^{-({false(1-{\mathrm{\omega }}_{P_t{P}_i}false)}^{2}false/(2{\mathrm{\sigma }}^2))},ffalse({\mathrm{\omega }}_{D_j{D}_k}false)=e^{-({false(1-{\mathrm{\omega }}_{D_j{D}_k}false)}^{2}false/(2{\mathrm{normal\sigma }}^2))}$, and σ was a parameter controlling the rate of variation in edge weight.…”