Thompson Corbin G, Sedykh Alexander, Nicol Melanie R, Muratov Eugene, Fourches Denis, Tropsha Alexander, Kashuba Angela D M
1 Division of Pharmacotherapy and Experimental Therapeutics, University of North Carolina Eshelman School of Pharmacy , Chapel Hill, North Carolina.
AIDS Res Hum Retroviruses. 2014 Nov;30(11):1058-64. doi: 10.1089/AID.2013.0254. Epub 2014 Mar 13.
The exposure of oral antiretroviral (ARV) drugs in the female genital tract (FGT) is variable and almost unpredictable. Identifying an efficient method to find compounds with high tissue penetration would streamline the development of regimens for both HIV preexposure prophylaxis and viral reservoir targeting. Here we describe the cheminformatics investigation of diverse drugs with known FGT penetration using cluster analysis and quantitative structure-activity relationships (QSAR) modeling. A literature search over the 1950-2012 period identified 58 compounds (including 21 ARVs and representing 13 drug classes) associated with their actual concentration data for cervical or vaginal tissue, or cervicovaginal fluid. Cluster analysis revealed significant trends in the penetrative ability for certain chemotypes. QSAR models to predict genital tract concentrations normalized to blood plasma concentrations were developed with two machine learning techniques utilizing drugs' molecular descriptors and pharmacokinetic parameters as inputs. The QSAR model with the highest predictive accuracy had R(2)test=0.47. High volume of distribution, high MRP1 substrate probability, and low MRP4 substrate probability were associated with FGT concentrations ≥1.5-fold plasma concentrations. However, due to the limited FGT data available, prediction performances of all models were low. Despite this limitation, we were able to support our findings by correctly predicting the penetration class of rilpivirine and dolutegravir. With more data to enrich the models, we believe these methods could potentially enhance the current approach of clinical testing.
口服抗逆转录病毒（ARV）药物在女性生殖道（FGT）中的暴露情况各不相同，且几乎无法预测。确定一种有效的方法来寻找具有高组织穿透力的化合物，将简化HIV暴露前预防和靶向病毒储存库方案的开发。在此，我们描述了使用聚类分析和定量构效关系（QSAR）建模对具有已知FGT穿透力的多种药物进行的化学信息学研究。对1950年至2012年期间的文献检索确定了58种化合物（包括21种ARV，代表13类药物），并给出了它们在宫颈或阴道组织或宫颈阴道液中的实际浓度数据。聚类分析揭示了某些化学类型在穿透能力方面的显著趋势。利用药物的分子描述符和药代动力学参数作为输入，采用两种机器学习技术建立了预测归一化至血浆浓度的生殖道浓度的QSAR模型。预测准确性最高的QSAR模型的R(2)test = 0.47。高分布容积、高MRP1底物概率和低MRP4底物概率与FGT浓度≥血浆浓度的1.5倍相关。然而，由于可用的FGT数据有限，所有模型的预测性能都很低。尽管有此限制，我们通过正确预测rilpivirine和dolutegravir的穿透类别，证实了我们的研究结果。有了更多数据来丰富模型，我们相信这些方法可能会改进当前临床测试的方法。
