Drug response prediction by ensemble learning and drug-induced gene expression signatures

Abstract

Chemotherapeutic response of cancer cells to a given compound is one of the most fundamental information one requires to design anti-cancer drugs. Recently, considerable amount of drug-induced gene expression data has become publicly available, in addition to cytotoxicity databases. These large sets of data provided an opportunity to apply machine learning methods to predict drug activity. However, due to the complexity of cancer drug mechanisms, none of the existing methods is perfect. In this paper, we propose a novel ensemble learning method to predict drug response. In addition, we attempt to use the drug screen data together with two novel signatures produced from the drug-induced gene expression profiles of cancer cell lines. Finally, we evaluate predictions by in vitro experiments in addition to the tests on data sets. The predictions of the methods, the signatures and the software are available from http://mtan.etu.edu.tr/drug-response-prediction/.

Toward Explainable Anticancer Compound Sensitivity Prediction via Multimodal Attention-Based Convolutional Encoders

“…There have been a plethora of works on the prediction of drug sensitivity in cancer cells (Garnett et al, 2012;Yang et al, 2012;Costello et al, 2014;Ali & Aittokallio, 2018;Kalamara et al, 2018). While the majority of them have focused on the analysis of unimodal datasets (genomics or transcriptomics, e.g., De Niz et al (2016); Tan (2016); Turki & Wei (2017); Tan et al (2018)), a handful of pre-vious works have integrated omics and chemical descriptors to predict cell line-drug sensitivity using a variety of methods including but not limited to: simple neural networks (one hidden layer) and random forests (Menden et al, 2013), kernelized Bayesian matrix factorization (Ammad- Ud-Din et al, 2014), Pearson correlation-based similarity networks (Zhang et al, 2015), a Kronecker product kernel in conjunction with SVMs (Wang et al, 2016), autoencoders in combination with elastic net and SVMs (Ding et al, 2018), matrix factorization (Wang et al, 2017), trace norm regularization (Yuan et al, 2016), link predictions (Stanfield et al, 2017) and collaborative filtering (Liu et al, 2018;Zhang et al, 2018b). In addition to genomic and chemical features, previous studies have demonstrated the value of complementing drug sensitivity prediction models with prior knowledge in the form of protein-protein interactions (PPI) networks (Oskooei et al, 2018b).…”