Machine Learning Approaches for the Identification of Genetic Interactions

Abstract

Genetic interactions are crucial in understanding the crosstalk between the gene pairs and help decipher their functional roles. They are defined as phenotypic outcomes resulting from two or more gene interactions. A gene proves to be an exceptional drug target of which the partner gene is mutated or overexpressed. Genetic interaction in cancer has been widely used for targeted therapy, of which synthetic lethality (SL) is the most studied. SL is a negative interaction in which inhibiting either of the genes does not affect the cell viability, but inhibiting both genes makes the cancer cell lethal. Various experimental and computational methods have been developed to identify and predict these interactions. This book chapter reviews different machine learning methods to predict SL interaction and how it mediates drug sensitivity. The chapter takes the reader through the salient features of different classical machine learning algorithms and their limitation. It also provides comprehensive knowledge about the features utilized for the model training and their importance. By the end of this book chapter, the readers will have an overview of different methods to identify genetic interaction and their associated advantages and limitations.

Keywords: Cancer; Genetic interactions; Machine learning; Synthetic lethality (SL).