Data for: RANKING STRATEGIES TO SUPPORT TOXICITY PREDICTION: A CASE STUDY ON POTENTIAL LXR BINDERS

A dataset of 356 compounds, mainly drugs or drug candidates, which consisted of groups of congeneric series sharing a common scaffold. The collected “LXR binders” covered a wide range of binding affinity, with IC50 values spanning from 1 nM to greater than 10000 nM. The dataset of LXR binders was enriched with decoy molecules, i.e. molecules that are presumed to be inactive against a target (they will not likely bind to the target). Decoys are commonly used to validate the performance of molecular modelling studies, as for example molecular docking, which was used in the present work. One-thousand decoy molecules were selected from Schrodinger 1K Drug-Like Ligand Decoys Set. For these molecules results obtained from the following modelling approaches are reported: ensemble docking, ePharmacophore, fingerprint similarity, structural alerts and QSAR PLS model. These results were used to build ranking strategies proposed in the paper.本数据集包含356种化合物，主要为药物或药物候选物，均属于共享共同母核的同系物系列。本次收集的“LXR结合剂（LXR binders）”涵盖了广泛的结合亲和力范围，其半数抑制浓度（IC50）跨度为1 nM至10000 nM以上。该LXR结合剂数据集还富集了诱饵分子——即被认为对靶标无活性（不太可能与靶标结合）的分子。诱饵分子常被用于验证分子建模研究的性能，例如本研究中使用的分子对接技术。本次研究从Schrodinger公司的1K类药配体诱饵数据集（Schrodinger 1K Drug-Like Ligand Decoys Set）中选取了1000个诱饵分子。针对这些分子，本研究报告了以下建模方法得到的结果：集成对接（ensemble docking）、电子药效团（ePharmacophore）、指纹相似度分析（fingerprint similarity）、结构警示（structural alerts）以及定量构效关系偏最小二乘模型（QSAR PLS model）。上述结果被用于构建本文提出的排序策略。