PiLSL is a pairwise interaction learning-based graph neural network (GNN) model for prediction of synthetic lethality (SL) as anti-cancer drug targets. It learns the representation of pairwise interaction between two genes from a knowledge graph (KG).