Comparative Modeling and Benchmarking Data Sets for Human Histone Deacetylases and Sirtuin Families

Jie Xia, Ermias Lemma Tilahun, Eyob Hailu Kebede, Terry-Elinor Reid, Liangren Zhang, Xiang Simon Wang
<span title="2015-02-09">2015</span> <i title="American Chemical Society (ACS)"> <a target="_blank" rel="noopener" href="https://fatcat.wiki/container/uilc246nrbhffhjy3flluhazbi" style="color: black;">Journal of Chemical Information and Modeling</a> </i> &nbsp;
Histone Deacetylases (HDACs) are an important class of drug targets for the treatment of cancers, neurodegenerative diseases and other types of diseases. Virtual screening (VS) has become fairly effective approaches for drug discovery of novel and highly selective Histone Deacetylases Inhibitors (HDACIs). To facilitate the process, we constructed the Maximal Unbiased Benchmarking Data Sets for HDACs (MUBD-HDACs) using our recently published methods that were originally developed for building
more &raquo; ... iased benchmarking sets for ligand-based virtual screening (LBVS). The MUBD-HDACs covers all 4 Classes including Class III (Sirtuins family) and 14 HDACs isoforms, composed of 631 inhibitors and 24,609 unbiased decoys. Its ligand sets have been validated extensively as chemically diverse, while the decoy sets were shown to be property-matching with ligands and maximal unbiased in terms of "artificial enrichment" and "analogue bias". We also conducted comparative studies with DUD-E and DEKOIS 2.0 sets against HDAC2 and HDAC8 targets, and demonstrate that our MUBD-HDACs is unique in that it can be applied unbiasedly to both LBVS and SBVS approaches. In addition, we defined a novel metric, i.e. NLBScore, to detect the "2D bias" and "LBVS favorable" effect within the benchmarking sets. In summary, MUBD-HDACs is the only comprehensive and maximalunbiased benchmark data sets for HDACs (including Sirtuins) that is available so far. MUBD-HDACs is freely available at http://www.xswlab.org/. Supporting Information All the data sets of MUBD-HDACs are freely accessible at http://www.xswlab.org. In addition, the Supporting Information (SI) contains the number of "duplicate" ligands, the activity types/cutoff for each HDACs isoform, the examples of "duplicate" ligands, the mean(ROC AUCs) values from Leave-One-Out Cross-Validation based on "simp"-based similarity search and MACCS "sims"-based similarity search (i.e. data for Figure 4) , top five pairs of structurally similar ligands/decoys for HDAC2 and HDAC8 among MUBD-HDACs, DUD-E and DEKOIS 2.0. This material is available free of charge via the Internet at Ligand Enrichment-The ROC curves calculated from molecular docking by GOLD are shown in Figure 6 (A), while their values of ROC AUC are also listed in Table 2 . For both targets of HDAC2 and HDAC8, the ROC curves for MUBD-HDACs are closer to the random distribution curve (diagonal line) than other two databases. Consistently, the ROC AUCs for MUBD-HDACs are 0.623 for HDAC2 and 0.618 for HDAC8, much less than those values with DUD-E of 0.722 for HDAC2 & 0.757 for HDAC8, and DEKOIS 2.0 of 0.762 for HDAC2 & 0.754 for HDAC8. These results indicate that the ligands in MUBD-HDACs are more challenging to be enriched from the background of our unbiased decoys by docking with GOLD. There seems to be no significant difference between DUD-E and DEKOIS 2.0 in this regard. Figure 6 (B) shows the ROC curves generated from similarity search based on FCFP_6; Table 2 lists the values of mean(ROC AUCs)s of those curves for different data sets. Similarly, for both targets the ROC curves for MUBD-HDACs are the closest to random distribution curves, followed by DUD-E. DEKOIS 2.0 seems to be away from the diagonal lines somehow. The values of mean(ROC AUCs)s show the same trend to ROC curves. For both HDAC2 and HDAC8, the values are the lowest in MUBD-HDACs (e.g. 0.666 for HDAC8) and the highest in DEKOIS 2.0 (e.g. 0.903 for HDAC8) while DUD-E (e.g. 0.831 for HDAC8) ranks in the middle. These curves and values indicate that MUBD-HDACs data sets are more challenging than the other two benchmarking sets for similarity search using FCFP_6 fingerprint. In summary, the data sets of MUBD-HDACs for HDAC2 and HDAC8 are (1) as diverse as DUD-E while more diverse than DEKOIS 2.0; (2) comparable to DUD-E in property matching but better than DEKOIS 2.0; (3) more challenging and unbiased than both DUD-E and DEKOIS 2.0 for enrichment by both types of VS approaches, i.e. docking using GOLD and similarity search using FCFP_6 fingerprint. Higher Similarity of Decoys to Ligands in MUBD-HDACs-The Tc values based on FCFP_6 fingerprint, which measures structural similarities, were calculated between all ligands and decoys in MUBD-HDACs, DUD-E and DEKOIS 2.0. The distribution curves of Tc were plotted for the purpose of comparison (cf. Figure 7) . In fact, the curves for MUBD-HDACs shift to the right direction (i.e. higher similarity) for both HDAC2 and HDAC8. DUD-E's curves are located in the middle, while DEKOIS 2.0 is to the left. These data indicate the higher similarities of decoys to ligands in MUBD-HDACs followed by DUD-E and DEKOIS 2.0. In addition, the top five pairs of ligands and decoys based on the rank of Tc values were listed for three benchmarking sets in Tables S5 and S6. In those structures, we can observe higher resemblance between ligands and decoys in MUBD-HDACs compared to other two sets. In DUD-E, the most dissimilar compounds based on ECFP4 fingerprints were chosen as decoys, while DEKOIS 2.0 applied a stricter standard to select structurally divergent decoys, i.e. "avoidance of latent actives in the decoy set (LADS) score". Though both methods were valuable to avoid "false negatives", it may artificially increase the ligand enrichment. 45 In contrast, our method of selecting decoys for HDACs Xia et al.
<span class="external-identifiers"> <a target="_blank" rel="external noopener noreferrer" href="https://doi.org/10.1021/ci5005515">doi:10.1021/ci5005515</a> <a target="_blank" rel="external noopener" href="https://www.ncbi.nlm.nih.gov/pubmed/25633490">pmid:25633490</a> <a target="_blank" rel="external noopener" href="https://pubmed.ncbi.nlm.nih.gov/PMC4677826/">pmcid:PMC4677826</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/tfj3rsugs5eannty6k66f5ruei">fatcat:tfj3rsugs5eannty6k66f5ruei</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20200206040558/http://europepmc.org/backend/ptpmcrender.fcgi?accid=PMC4677826&amp;blobtype=pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/84/96/8496f344c601dfe9b7b5d8465199f1e0f156b5ba.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener noreferrer" href="https://doi.org/10.1021/ci5005515"> <button class="ui left aligned compact blue labeled icon button serp-button"> <i class="external alternate icon"></i> acs.org </button> </a> <a target="_blank" rel="external noopener" href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4677826" title="pubmed link"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> pubmed.gov </button> </a>