Pathway-Guided Deep Neural Network toward Interpretable and Predictive Modeling of Drug Sensitivity

Abstract

To efficiently save cost and reduce risk in drug research and development, there is a pressing demand to develop in silico methods to predict drug sensitivity to cancer cells. With the exponentially increasing number of multi-omics data derived from high-throughput techniques, machine learning-based methods have been applied to the prediction of drug sensitivities. However, these methods have drawbacks either in the interpretability of the mechanism of drug action or limited performance in modeling drug sensitivity. In this paper, we presented a pathway-guided deep neural network (DNN) model to predict the drug sensitivity in cancer cells. Biological pathways describe a group of molecules in a cell that collaborates to control various biological functions like cell proliferation and death, thereby abnormal function of pathways can result in disease. To take advantage of the excellent predictive ability of DNN and the biological knowledge of pathways, we reshaped the canonical DNN structure by incorporating a layer of pathway nodes and their connections to input gene nodes, which makes the DNN model more interpretable and predictive compared to canonical DNN. We have conducted extensive performance evaluations on multiple independent drug sensitivity data sets and demonstrated that our model significantly outperformed the canonical DNN model and eight other classical regression models. Most importantly, we observed a remarkable activity decrease in disease-related pathway nodes during forward propagation upon inputs of drug targets, which implicitly corresponds to the inhibition effect of disease-related pathways induced by drug treatment on cancer cells. Our empirical experiments showed that our method achieves pharmacological interpretability and predictive ability in modeling drug sensitivity in cancer cells. The web server, the processed data sets, and source codes for reproducing our work are available at .

Integration strategies of multi-omics data for machine learning analysis

“…As several methods have been developed for interpreting neural networks models, we invite our reader to other reviews [71] , [72] . Most of those strategies focus on explaining the final decision of the algorithm and identifying biomarkers, but some DL models [73] , [74] can directly find relevant biological pathways during the learning process ( Fig. 1 ).…”

“…Hence, important pathways implicated in the outcome are activated with bigger weights during training. Figure inspired from Deng et al (2020) [73] . …”

Multiple Laplacian Regularized RBF Neural Network for Assessing Dry Weight of Patients With End-Stage Renal Disease

“…Deep learning ( Dao et al, 2021 ; Lv et al, 2021a , b ) also made a great contribution to the clinic, including skin cancer ( Esteva et al, 2017 ), breast cancer ( Liu J. et al, 2021 ), and brain diseases ( Liu G. et al, 2018 ; Liu et al, 2019 ; Bi et al, 2020 ; Hu et al, 2020 , 2021a , b ). In biological field, machine learning has been widely used to solve biological problems, including O -GlcNAcylation site prediction ( Jia et al, 2018 ), microbiology analysis ( Qu et al, 2019 ), microRNAs and cancer association prediction ( Zeng et al, 2018 ), lncRNAs ( Cheng et al, 2016 ; Deng et al, 2021 ), CircRNAs ( Fang et al, 2019 ; Zhao et al, 2019 ), DNA methylation site ( Wei et al, 2018b ; Zou et al, 2019 ; Dai et al, 2020 ), osteoporosis diagnoses ( Su et al, 2020b ), function prediction of proteins ( Wei et al, 2018a ; Wang H. et al, 2019 ; Deng et al, 2020b ; Ding et al, 2020a ; Su et al, 2020a ), nucleotide binding sites ( Ding et al, 2021b ), drug complex network analysis ( Ding et al, 2019 , 2020b , a ; Deng et al, 2020a ; Han et al, 2021 ; Liu H. et al, 2021 ), protein remote homology ( Liu B. et al, 2018 ), electron transport proteins ( Ru et al, 2019 ), and cell-specific replication.…”

Machine learning in the prediction of cancer therapy

“…Initially, feedforward deep neural networks (DNNs) were applied to develop models using selected genomic features [80] or transcriptomic data [88] . Later studies incorporated selected gene expression features with pathway information to build DNN models [89] , [90] . In any case, all these DNN models have been shown to outperform classical ML models.…”