A context-based matched molecular pair analysis identifies structural transformations that reduce CYP1A2 inhibition.

Cytochrome P450 1A2 (CYP1A2) metabolizes ∼10-15% of FDA-approved drugs. Available quantitative structure-activity relationship (QSAR) and machine learning methods offer little design insights to reduce CYP1A2 inhibition. We performed matched molecular pair analysis (MMPA) on the CYP1A2 inhibition dataset (ChEMBL3356) and identified key structural transformations. A chemical context-based analysis was performed using Kramer's method to tackle the limitations of the global MMPA. The global MMPA agreed with earlier QSAR studies (influence of H to F, Me, OMe, and OH transformations). Our results show that the effect of these transformations depends on the local chemical environment. The H to Me transformation reduced the inhibition in three pharmacologically important scaffolds (e.g., indanylpyridine). Structure-based analysis (docking) showed that the interaction of the heteroatoms with Heme-Fe is influenced by useful transformations. Overall, this work presents the first context-based analysis of the CYP1A2 dataset and offers novel medicinal chemistry insights useful for lead optimization.