Assessment of quantum mechanics-based hydropathic descriptors to navigate the isofunctional chemical space in scaffold hopping

Scaffold hopping is a well-established paradigm in modern drug discovery, providing valuable support in the hit-to-lead and lead optimization stages due to the impact in (i) improving the activity via the exploration of structure–activity relationships, (ii) circumventing intellectual property restrictions, and (iii) optimizing the pharmacokinetics profile of drug-like candidates. The effectiveness of this approach, however, critically depends on the quality of the parameters used to describe chemical scaffolds, as appropriate filtering and prioritization can be decisive for the success of scaffold hopping. In this context, quantum mechanical (QM)-based methods have demonstrated superior performance in capturing the physicochemical features of small molecules, making them a robust complementary option for 3D similarity analyses and virtual screenings. Here we evaluated the suitability of the QM-based hydropathic (HyPhar) descriptors derived from continuum solvation calculations for scaffold hopping and navigating the isofunctional fragment space. To this end, a representative subset of real-case scaffold-hopping examples, which was assembled from the ‘large-hops’ (LH) benchmark dataset, was used to validate our methodology. The results showed that the HyPhar descriptors performed well in the identification of the active scaffold, outperforming the behavior observed for classical 2D and 3D molecular representations. In addition, these descriptors were also used to examine the ability to discriminate active scaffolds from decoys exploiting the hydropathicity-guided alignment for five distinct targets taken among the LH benchmark, mimicking the scenario faced in drug discovery studies. Notably, a higher early enrichment was obtained relative to docking-based computations, thus demonstrating superior capability in discriminating active scaffolds from decoys. In particularly, we proposed the use of HyPhar descriptors to rescore the poses of docked scaffolds due to the overall improvement in the prioritization of the isofunctional fragments relative to the discrimination obtained from docking calculations. Scientific Contribution Time-efficient and accurate in silico approaches are critical for the success of drug discovery programs. Here, we demonstrate that QM-MST HyPhar descriptors could provide a rapid and effective strategy for filtering and prioritizing isofunctional scaffolds in scaffold hopping. Notably, the results indicate better performances compared to other commonly used 2D, 3D scoring metrics and machine learning-based chemogenomic methods.