Collaborative filtering on a family of biological targets

COMP 214

Dumitru Erhan, erhandum@iro.umontreal.ca1, Pierre-Jean L'Heureux, lheureup@iro.umontreal.ca1, Yoshua Bengio, bengioy@iro.umontreal.ca1, and Shi Yi Yue, ShiYi.Yue@astrazeneca.com2. (1) Department IRO, Université de Montréal, Montreal, QC H3C 3J7, Canada, (2) Department of Chemistry, AstraZeneca R&D Montreal, 7171 Frederick-Banting, St-Laurent, QC H4S 1Z9, Canada
Building a QSAR model of a new biological target for which few screening data is available is a daunting task. However, the new target may be part of a bigger family, for which we have more screening data. Collaborative filtering is a field of machine learning that tries to build predictive models that link multiple targets to multiple examples. If there are more commonalities between the targets, a better multi-target model can be built. We show an example of a multi-target neural network that can use family information to produce a predictive model of an undersampled target. We show its performance on compound prioritization for an HTS campaign, and the underlying shared representation between targets.