Construction of a novel scoring function by using a genetic algorithm for virtual screening of histone deacetylase inhibitors (HDACi)

The comparison of different scoring functions on the discrimination and enrichment ability of histone deacetylase II (HDAC2) inhibitors were explored. A self-constructed database containing 45 inhibitors of HDAC2 and 2285 inactive compounds was performed screening study by using Surflex-Dock. The discrimination and enrichment ability of five scoring functions, total_score D_score, PMF_score, G_score and Chemscore, were evaluated by means of receiver operating characteristic (ROC) and enrichment factor (EF). A novel scoring model S_combine with different coefficients was constructed by using the five basic scoring functions. The coefficients were determined via a genetic algorithm. According to the performance of ROC and EF calculations, the new scoring term S_combine proposed exhibited superiority over the conventional scoring methods, which improved the accuracy of virtual screening for HDAC2 inhibitors.