Kramer, Christian and Gedeck, Peter (2010) Leave-cluster-out crossvalidation is appropriate for Scoring Functions derived on diverse protein datasets. Journal of Chemical Information and Modeling, 50 (11). pp. 1961-1969. ISSN 1549-960X

Abstract

With the emergence of large collections of protein-ligand complexes complemented by binding data, as found in PDBbind or BindingMOAD, new opportunities for parameterizing and evaluating scoring functions arise. With huge data collections available it becomes feasible to fit scoring functions in a QSAR style, i.e. by defining protein-ligand interaction descriptors and analyzing them with modern machine-learning methods. As in each data modelling ansatz, care has to be taken to validate the model carefully. Here we show that there are large differences measured in R (0.77 vs. 0.46) or R2 (0.59 vs. 0.21) for a relatively simple scoring function depending on whether it is validated against the PDBbind core set or validated in a leave-cluster-out crossvalidation. If proteins from the same family are present in both training and validation set, the estimated prediction quality from standard validation techniques looks too optimistic.

Item Type:	Article
Related URLs:	http://dx.doi.org/10.1021/ci100264e
Related URLs:	http://dx.doi.org/10.1021/ci100264e
Date Deposited:	13 Oct 2015 13:16
Last Modified:	13 Oct 2015 13:16
URI:	https://oak.novartis.com/id/eprint/3140