Abstract

Genes whose function is selectively essential in the presence of cancer driver gene alterations represent promising targets for the development of precision therapeutics. Over the last decade multiple groups have performed large-scale loss-of-function screens in panels of cancer cell lines in order to identify such genetic dependencies. Despite these systematic efforts, relatively few robust genetic dependencies have been identified, with many 'hits' appearing to be cell line or screen specific.  We have previously used protein interaction networks as a means of interpreting the genetic dependencies identified in loss-of-function screens. Here, we show that genetic dependencies involving pairs of genes whose protein products interact are more likely to be reproduced across multiple experiments. Thus the integration of protein-protein interaction networks serves as both a means of interpreting genetic dependencies and a way to identify those genetic dependencies that are likely to be reproduced across multiple experiments.