Application of Transcriptomics for Predicting Protein Interaction Networks, Drug Targets and Drug Candidates

Protein interaction pathways and networks are critically-required for a vast range of biological processes. Improved discovery of candidate druggable proteins within specific cell, tissue and disease contexts will aid development of new treatments. Predicting protein interaction networks from gene expression data can provide valuable insights into normal and disease biology. For example, the resulting protein networks can be used to identify potentially druggable targets and drug candidates for testing in cell and animal disease models. The advent of whole-transcriptome expression profiling techniques—that catalogue protein-coding genes expressed within cells and tissues—has enabled development of individual algorithms for particular tasks. For example,: (i) gene ontology algorithms that predict gene/protein subsets involved in related cell processes; (ii) algorithms that predict intracellular protein interaction pathways; and (iii) algorithms that correlate druggable protein targets with known drugs and/or drug candidates. This review examines approaches, advantages and disadvantages of existing gene expression, gene ontology, and protein network prediction algorithms. Using this framework, we examine current efforts to combine these algorithms into pipelines to enable identification of druggable targets, and associated known drugs, using gene expression datasets. In doing so, new opportunities are identified for development of powerful algorithm pipelines, suitable for wide use by non-bioinformaticians, that can predict protein interaction networks, druggable proteins, and related drugs from user gene expression datasets.