Identification of structural fingerprints for ABCG2 inhibition by using Monte Carlo optimization, Bayesian classification, and structural and physicochemical interpretation (SPCI) analysis

• Published in: SAR and QSAR in environmental research Vol. 31; no. 6; pp. 439 - 455
• Main Authors: Ghosh, K., Bhardwaj, B., Amin, S.A., Jha, T., Gayen, S.
• Format: Journal Article
• Language: English
• Published: Abingdon Taylor & Francis 02.06.2020; Taylor & Francis Ltd
• Subjects: ABCG2; Bayesian analysis; Bayesian classification model; Breast cancer; Breast cancer resistance protein (BCRP); Chemotherapy; Classification; Computer simulation; Fingerprints; machine learning; Mathematical models; Molecular structure; Monte Carlo optimization; Monte Carlo simulation; Multidrug resistance; Optimization; QSAR; Statistical analysis; Statistical methods; Structure-activity relationships
• ISSN: 1062-936X; 1029-046X
• DOI: 10.1080/1062936X.2020.1771769

The human breast cancer resistance protein (BCRP), one of the members of the large ATP binding cassette (ABC) transporter superfamily, is crucial for resistance against chemotherapeutic agents. Currently, it has been emerged as one of the best biological targets for the designing of small molecule drugs capable of eliminating multidrug resistance in breast cancer. In order to gain insights into the relationship between the molecular structure of compounds and the ABCG2 inhibition, a multi-QSAR approach using different methods was performed on a dataset of 294 ABCG2 inhibitors with diverse scaffolds. The best models obtained by different chemometric methods have the following statistical characteristics: Monte Carlo Optimization-based QSAR (sensitivity = 0.905, specificity = 0.6255, accuracy = 0.756, and MCC = 0.545), Bayesian classification model (sensitivity = 0.735, specificity = 0.775, and concordance = 0.757); structural and physicochemical interpretation analysis-random forest method (balance accuracy = 0.750, sensitivity = 0.810, and specificity = 0.700). Additionally, structural fingerprints modulating the ABCG2 inhibitory properties were identified from the best models of each method and also validated with each other. The current modelling study is an attempt to get a deep insight into the different important structural fingerprints modulating ABCG2 inhibition.