Scaffold Hopping Using Two-Dimensional Fingerprints: True Potential, Black Magic, or a Hopeless Endeavor? Guidelines for Virtual Screening

Abstract

The scaffold hopping potential of popular 2D fingerprints has been thoroughly investigated. We have found that these types of fingerprints have at least limited scaffold hopping ability including early enrichment of small numbers of active scaffolds at high database ranks. However, it has not been possible to derive Tanimoto coefficient value ranges for individual fingerprints that are generally preferred for scaffold hopping. For selected fingerprints, similarity threshold values have been identified that yield small database selection sets having a high probability to contain a few active scaffolds. Furthermore, essentially all tested fingerprints have shown the ability to enrich scaffold hops in approximately 1% of a screening database. For the test cases reported herein, selecting 0.5-1% of the screening database yields approximately 25% of the available scaffolds. On the basis of our findings, practical guidelines for virtual screening using different types of 2D fingerprints have been formulated.

Effectiveness of 2D Fingerprints for Scaffold Hopping

“…The MUV searches have been excluded from Table 5 as the W value of 0.03 was not statistically significant at the 0.01 level, i.e., there was no degree of consistency in the ranking of the different fingerprints. Similar comments apply to the MDDR and WOMBAT datasets if attention is restricted to just those activity classes with mean similarities < 0.40 in Tables 1a and 1b. While this article was being prepared for publication, we became aware of a very recent report by Vogt et al that is closely related to our work [27]. They used sets of known actives for 17 biological targets, adding these to ZINC and ChEMBL datasets each containing ca.…”

Systematic assessment of scaffold hopping versus activity cliff formation across bioactive compound classes following a molecular hierarchy

“…Therefore, through the application of computational methods, for example, pharmacophore or whole-molecule similarity search, it is attempted to extrapolate from known active compounds and identify others that share the same activity but are structurally distinct. [7][8][9] As such, scaffold hopping is typically considered the ultimate goal of virtual screening campaigns 7-10 and of highthroughput screening data analysis. 11 Through large-scale compound data mining, it has been shown that many target-specific compound activity classes contain large numbers of different scaffolds 12 and that diverse scaffolds often represent specifically active compounds with comparably high potency.…”

Follow up - Compound data sets and software tools for chemoinformatics and medicinal chemistry applications: update and data transfer

“…These MMPs consider chemical reaction information and should be useful for practical medicinal chemistry applications. [18] 2009 VS_ML_5 20 AC from the literature and 15 AC from the Molecular Drug Data Report 12 [19] 2010 VS_ML_6 8 AC 13 [20] 2010 PROG_3 Program to generate target selectivity patterns of scaffolds 14 [21] 2010 PROG_4 Multi-target CAGs (see also entry 10) with a sample set containing 33 kinase inhibitors 15 [22] 2010 PROG_5 SARANEA 16 [23] 2010 PROG_6 3D activity landscape program with a sample set containing 248 cathepsin S inhibitors 17 [24] 2010 SAR_1 2 sets of MMPs from BindingDB and ChEMBL 18 [25] 2010 PROG_7 Similarity-potency tree (SPT) program with a sample set containing 874 factor Xa inhibitors 19 [26] 2010 VS_ML_7 17 target-directed compound sets; each set contains a minimum of 10 distinct scaffolds and each scaffold represents 5 compounds 20 [27] 2011 SAR_VZ 10,489 malaria screening hits 21 [28] 2011 SAR_2 458 target-based sets with scaffolds and scaffold hierarchies 22 [29] 2011 SAR_VZ 4 sets of compounds active against 3 or 4 targets 23 [30] 2011 SAR_VZ 881 factor Xa inhibitors 24 [31] 2011 VS_ML_8 50 AC prioritized for similarity searching 25 [32] 2011 VS_ML_9 25 data sets from successful ligand-based virtual screening applications 26 [33] Data entries are organized according to scientific subject areas: structure-activity relationship (SAR) and structure-selectivity relationship (SSR) analysis, SAR visualization (SAR_VZ), virtual screening via similarity searching or machine learning (VS_ML), and programs (PROG). References in the Entry column provide the original publication introducing the program and/or data set.…”