Set 76 Benzothiophene-3- boronic acid 69 186 17 100The distribution of data set for every HDAC isoform is unbalanced. Therefore, we divided the 1-phenyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-1H-pyrazole data set into a training set (65 ) and a test set (35 ) by stratified sampling [38] (Additional file 2: Table S1, Additional file 3: Table S2).HDAC6/ligands HDAC7/ligands HDAC8/ligands TotalWu et al. BMC Bioinformatics 2012, 13:212 http://www.biomedcentral.com/1471-2105/13/Page 6 ofTable 9 11 sequence similarity descriptors of HDAC2, 4, 6, 7 andClass I HDAC1 Class I HDAC2 HDAC8 Class IIa HDAC4 HDAC7 Class IIb HDAC6 9.5 9.4 7.7 8.1 20.6 17.2 19.1 16.6 100.0 23.4 7.4 11.1 10.0 10.2 9.7 9.5 10.5 9.0 9.9 100.0 46.9 58.4 40.5 46.9 100.0 54.3 39.7 20.6 19.1 11.1 13.0 8.6 9.6 85.1 30.8 100.0 30.7 51.9 34.4 30.7 100.0 10.2 9.0 9.6 8.4 9.7 9.9 9.9 10.8 9.4 8.1 14.2 13.6 18.9 21.4 HDAC2 HDAC3 HDAC8 Class IIa HDAC4 HDAC5 HDAC7 HDAC9 Class IIb HDAC6 HDAC10 Class IV HDACFor all possible HDACs pairs, sequence identities (in ) are reported.length, bond angle and dihedral angle [42], 30 protein Geometry descriptors were obtained for each HDAC protein (Additional file 4: Table S3).Description of inhibitorsIn our study, the HDAC inhibitors were represented by two kinds of feature space, i.e. 32-dimensional General Descriptors (GD) and 28-dimensional Drug-Like Index (DLI). These descriptors are widely applied to the construction of QSAR models. For general descriptors, they include atomic contributions to van der waals surface area, log P (octanol/water), molar refractivity, PubMed ID:https://www.ncbi.nlm.nih.gov/pubmed/20618955 and partial charge. GD characterize physical properties and describe organic compounds in boiling point, vapor pressure, free energy of salvation in water, solubility in water, thrombin/trysin/ factor Xa activity, blood rain barrier permeability, and compound classification etc. [43]. On the other hand, DLI characterize simple topological indices of compounds and measure the hierarchy of drug structures in terms of rings, links, and molecular frameworks [44].Protein-inhibitor cross-termsdescriptors. In reality, protein-ligand interactions are governed by complex processes that depend on the complementarity of the properties of PubMed ID:https://www.ncbi.nlm.nih.gov/pubmed/12711626 the interacting entities. In PCM, this is accounted for by protein-inhibitor crossterms [31,36], which in the simplest case is obtained by multiplication of mean centered descriptors of proteins and inhibitors. Therefore, we obtained 11 ?32 = 352, 5 ?32 = 160, 30 ?32 = 960, 11 ?28 = 308, 5 ?28 = 140, 30 ?28 = 840 cross-terms for P0-GD, P1-GD, P2-GD, P0-DLI, P1-DLI, and P2-DLI respectively.Preprocessing of dataTo reduce the bias of the model, all descriptors were mean centered and scaled to unit variance prior to the calculation of protein-ligand cross-terms. Moreover, the binding affinities (IC50) were logarithmically transformed to pIC50 and then mean centered and scaled to unit variance.Proteochemometric modelingEvidently, ligand-receptor recognition can only be partially explained by linear combinations of ligand and receptorTable 10 Five protein structure similarity descriptors of HDAC2, 4, 6, 7 andClass I HDAC2 Class I HDAC2 HDAC8 Class IIa HDAC4 HDAC7 Class IIb HDAC6 0.031 0.048 0.027 0.036 1.000 0.182 0.182 0.180 0.186 1.000 0.706 0.706 1.000 0.027 0.036 1.000 0.407 0.407 1.000 0.182 0.180 0.182 0.186 0.031 0.048 HDAC8 Class IIa HDAC4 HDAC7 Class IIb HDACSupport vector machine (SVM) is a non-linear modeling technique applied multiple times in PCM [33,45-50]. We created PCM models using support vector regression (SVR) built in Weka suit (Weka.