Multiple search methods for similarity-based virtual screening: analysis of search overlap and precision

Abstract

BackgroundData fusion methods are widely used in virtual screening, and make the implicit assumption that the more often a molecule is retrieved in multiple similarity searches, the more likely it is to be active. This paper tests the correctness of this assumption.ResultsSets of 25 searches using either the same reference structure and 25 different similarity measures (similarity fusion) or 25 different reference structures and the same similarity measure (group fusion) show that large numbers of unique molecules are retrieved by just a single search, but that the numbers of unique molecules decrease very rapidly as more searches are considered. This rapid decrease is accompanied by a rapid increase in the fraction of those retrieved molecules that are active. There is an approximately log-log relationship between the numbers of different molecules retrieved and the number of searches carried out, and a rationale for this power-law behaviour is provided.ConclusionsUsing multiple searches provides a simple way of increasing the precision of a similarity search, and thus provides a justification for the use of data fusion methods in virtual screening.

eFindSite: Enhanced Fingerprint‐Based Virtual Screening Against Predicted Ligand Binding Sites in Protein Models

“…Nevertheless, the overall performance of data fusion techniques, particularly using the SUM rule, is superior compared to other methods. This is consistent with previous studies on multiple search methods showing a systematic improvement of compound ranking by applying data fusion techniques 68,69. Comparing results obtained for crystal structures to those for different quality protein models demonstrates a fairly high insensitivity of e FindSite to the structure deformations of target receptors.…”

Chemically Advanced Template Search (CATS) for Scaffold‐Hopping and Prospective Target Prediction for ‘Orphan’ Molecules

“…It might thus be advisable to select one method from each group for similarity searching and compare ranked results lists, e.g. by data fusion 23. We wish to point out that the grouping of methods depicted in Figure 3 should be treated with caution, as the dendrograms are likely to vary for other reference data sets and chemotype/target coverage.…”

Chemoinformatics at the University of Sheffield 2002–2014

“…A more successful analytic study was reported by Holliday et al. [45] They demonstrated that a power law distribution could be used to predict to a fair degree of accuracy the numbers of database structures common to one similarity search, to two similarity searches, to three etc. They also demonstrated that the proportion of actives increased rapidly in these sets of common structures: the probability of activity of a database structure hence increases in line with its frequency of retrieval in multiple similarity searches, thus providing a simple empirical justification for the use of fusion methods in virtual screening.…”