Figure 1:

Schematic illustration of Expression Dose Dependent Inferelator (EDDI). Using dependency screening and RNAseq data, our study uses machine learning methods to discriminate modes of dependency and allow for the identification of dosage-based dependency predictor genes (DDPs). DDPs gene expression patterns in dependent and non-dependent cancer cells were used to predict dependency to a given gene. Ensemble models corresponding to each gene dependency were constructed and evaluated. The resulting dependent-predictor pairs obtained across all trained models captured dosage-based dependencies which were used to construct networks for dependent and predictor genes. Our proposed methodology allows us to extract biological mechanisms of dependency.