Martin, Eric, Polyakov, Valery and Tian, Li (2017) Additional supplementary materials for approved article OAK ID 31984. Journal of chemical information and modeling.

Abstract

These are the exact public-domain ChEMBL data sets, training/test set splits and activity predictions from the paper. There is no internal Novartis data. (After a presentation at ACS, many people asked for them.) They were already approved and released to an informal collaborator (Steven Kearnes) at google, OAK ID 32130 & 31607. Profile-QSAR on our "realistic" test set has come to be seen as the method to beat by other machine-learning virtual screening algorithms, so we are offering this as the test set for other researchers to use as a bench mark.

Item Type:	Article
Date Deposited:	14 Sep 2017 00:45
Last Modified:	14 Sep 2017 00:45
URI:	https://oak.novartis.com/id/eprint/32745