Enhancing the Effectiveness of Similarity-Based Virtual Screening Using Nearest-Neighbor Information

Abstract

We test the hypothesis that fusing the outputs of similarity searches based on a single bioactive reference structure and on its nearest neighbors (of unknown activity) is more effective (in terms of numbers of high-ranked active structures) than a similarity search involving just the reference structure. This turbo similarity searching approach provides a simple way to enhance the effectiveness of simulated virtual screening searches of the MDL Drug Data Report database.

Similarity-based virtual screening using 2D fingerprints

“…However, the fact that the average recall does increase, even with 100 nearest neighbours, means that even these molecules continue to provide useful structural information. At still larger numbers of nearest neighbours (200 in our experiments with these data), performance does flatten-off and then starts to decrease [50].…”

“…Most recently, the work described above on VS using multiple reference structures has led us to devise a novel, but very simple, way of enhancing the effectiveness of similarity-based VS when just a single reference structure is available [50]. We refer to this approach as turbo similarity searching; a turbocharger increases the power of an engine by using the engine's exhaust gases, and turbo similarity searching seeks to increase the power of a search engine procedure by using the reference structure's nearest neighbours.…”

“…This paper summarises recent work at Sheffield, focusing on the effects of varying the similarity coefficient that is used for searching, and of combining multiple database searches [44][45][46][47][48][49][50][51].…”

Chemoinformatics at the University of Sheffield 2002–2014

“…The similar property principle would suggest that the nearest neighbours of the reference structure in a similarity search are likely to exhibit the same bioactivity; if we then assume that they actually do exhibit that activity then we can use them as pseudo-reference structures in a group fusion search, combining the rankings resulting from their use with that resulting from the initial reference structure. [38][39] Thus far, we have described data fusion as involving the combination of multiple rankings of a database to produce a single, fused ranking that is the output from a similarity search. Our initial studies used simple arithmetic fusion rules that had first been described for the combination of rankings in textual information retrieval systems [40] as exemplified in Table 2.…”

Analysis and use of fragment-occurrence data in similarity-based virtual screening

“…Table 1a lists the MDDR classes, which were selected in collaboration with the Novartis Institutes for BioMedical Research and which have been used in several previous studies of ligand-based virtual screening by both ourselves and others (e.g., [34][35][36][37][38]). Each row of the table contains an activity class, a short abbreviation of the name, the number of molecules belonging to the class, the number of distinct ring systems occurring in the set of active molecules for the class, and an indication of the class's diversity.…”