Data Science Day @ Columbia University

Synthetic lethality (SL) is an interaction where two nonessential genes lead to cellular inviability when knocked out simultaneously. Recent work suggests it may be useful in cancer therapy. It may also predict novel drug-drug interactions and suggest mechanisms of adverse drug reactions. Few SL interactions have been found in humans to date, but many are known in yeast. Here, we present Species INdependent TRAnslation (SINaTRA), an algorithm that predicts SL in translation between species using protein-protein interaction (PPI) networks and connectivity homology. We validate SINaTRA using S. cerevisiae and S. pombe and find that model performance improves significantly with the use of SINaTRA. In addition, SINaTRA outperforms other methods of predicting SL, including genetic homology. We then apply our model to the human PPI network. After genetic filtering using 1000 Genomes data, we provide 100 million human gene pairs with measures of their predicted synthetic lethality (SINaTRA scores)