Quantitative Structure Activity Relationship Analysis of Antiviral Activity of PF74 Type HIV-1 Capsid Protein Inhibitors by Simplex Representation of Molecular Structure

Human immunodeficiency virus type 1 (HIV-1) capsid protein (CA) plays essential roles in both early and late stages of HIV-1 replication and it is an important target for developing new therapeutic drugs. In the present study, quantitative structure–activity relationships (QSARs) models were developed for a series of 46 derivatives of PF74 CA inhibitors binding to the HIV1 capsid. The Partial least square (PLS) method was applied to obtain QSAR models based on simplex representation of molecular structure (SiRMS). Molecular docking studies were performed to find the optimal conformation of molecules within the binding site, and to compute 3D SiRMS descriptors. The study produced several statistically significant and acceptable QSAR models (R2 = 0.98, Q2 = 0.76–0.78, R2 predict = 0.84). A consensus model developed from the five excellent PLS QSAR models which was statistically validated for its predictive power using cross-validation, Y-randomization, and applicability domain evaluation procedures. The obtained results have satisfactory statistical parameters within acceptable range. The model interpreted by a reverse analysis identified the molecular fragments that promote and interfere with antiviral activity and the degree to which they influence activity. The influence of the physical–chemical properties of the molecules on antiviral activity was determined. It has been observed that antiviral activity is mainly governed by electrostatic characteristics of molecular structure and hydrophobic factors. The results obtained may further assist in the design of new CA inhibitor compounds and prediction of their activity.

1型人类免疫缺陷病毒（Human immunodeficiency virus type 1, HIV-1）衣壳蛋白（capsid protein, CA）在HIV-1复制的早期与后期阶段均发挥关键作用，同时也是新型治疗药物研发的重要靶点。本研究针对46种结合HIV-1衣壳的PF74类CA抑制剂衍生物，构建了定量构效关系（quantitative structure–activity relationships, QSARs）模型。本研究基于分子结构单纯形表示法（simplex representation of molecular structure, SiRMS），采用偏最小二乘法（Partial least square, PLS）构建QSAR模型；并开展分子对接（molecular docking）实验，以确定配体在结合口袋内的最优构象，同时计算三维SiRMS描述符。本研究得到了多个统计学意义显著且性能优良的QSAR模型，其拟合系数R²为0.98，交叉验证系数Q²为0.76~0.78，预测系数R²predict为0.84。本研究基于5个性能优异的PLS-QSAR模型构建了集成模型，并通过交叉验证（cross-validation）、Y随机化检验（Y-randomization）以及适用域（applicability domain）评估流程，对该集成模型的预测能力进行了统计学验证。所得结果的统计学参数均处于可接受范围内，表现令人满意。通过反向分析对构建的模型进行解析，本研究明确了对抗病毒活性具有促进或抑制作用的分子片段，以及这些片段对活性的影响程度。本研究还明确了分子的理化性质对抗病毒活性的影响。研究发现，抗病毒活性主要由分子结构的静电特性与疏水因素决定。本研究所得结果可为新型CA抑制剂类化合物的设计及其活性预测提供参考。