Machine learning algorithms to predict drug responses

A. Schematic representation of the study design and bioinformatics pipeline. (a) Dataset: the full data set was constructed using the GDSC and CCLP databases (see also Supplemental Fig. 1 in ref 1). (b) Model construction: Association Rule Mining (ARM) was used to generate testable hypotheses of genes associated with sensitivity or resistance to specific drugs (left panel). (c) Validation: our models were validated computationally and in a variety of in vitro experimental settings.

B. A machine learning workflow was designed to predict drug response and survival of cancer patients. All pipelines were trained on a large panel of cancer cell lines and tested in clinical cohorts. Deep Neural Networks outperformed other machine learning algorithms and captured pathways that link gene expression with drug response (2).

1) Vougas et al. Machine learning and data mining frameworks for predicting drug response in cancer: An overview and a novel in silico screening process based on association rule mining. Pharmacol Ther 2019, 203: 107395.

2) Sakellaropoulos et al. A deep learning framework for predicting response to therapy in cancer. Cell Rep 2019, 29(11): 3367-3373.e4.

VG New

Prof. Vassilis G. Gorgoulis

Laboratory of Histology-Embryology
Molecular Carcinogenesis Group
Medical School
National and Kapodistrian University of Athens


Biomedical Research Foundation of the Academy of Athens


Faculty Institute for Cancer Sciences, University of Manchester,
Manchester Academic Health Science Centre, Manchester, UK

Manchester Centre for Cellular Metabolism,
University of Manchester, Manchester Academic Health Science Centre, Manchester


EMBO member






Office Tel: 0030 210-7462352
Fax: 0030 210-7462340

Also see...


Copyright ©2011-2015, Prof. Gorgoulis Powered by AVMap