Pharmacophore-driven kinase profiling applied to the PKIS2 chemogenomic dataset

Abstract We present a novel and efficient approach for extracting 3D pharmacophores, without any supervised compound selection, from a chemogenomic kinase dataset (406 kinases and 645 compounds, PKIS2) using the NORNS software. For each pharmacophore, we introduce a metric called NEM, which quantifies the change in the proportion of active compounds consistent with the pharmacophore compared to the original dataset. Based on this metric, we can identify pharmacophores associated with specific kinases and, conversely, determine all kinases that share similar metric values. This approach enables the characterization of polypharmacological profiles linked to individual pharmacophores.The consistency of our results with various biological datasets, including ChEMBL, DrugBank, LINCS, KINOMEscan, and Kinobeads, was evaluated and showed strong correlations in selected case studies. This study highlights the potential of our approach for elucidating relationships between pharmacophores and kinase selectivity profiles, offering valuable insights into kinase–target interactions.Scientific contributionStarting from our previous software, NORNS (see Availability of data and materials for the current version), which focused on 2D pharmacophores extracted from a global chemical dataset and associated with a single biological profile, this article now explores 3D pharmacophores across multiple conformations, linked to a polypharmacological profile with a focus on kinases. We demonstrate the ability to extract pharmacophores associated with a specific kinase, and define a selectivity profile for these pharmacophores with respect to other kinases. We also show the possibility of extracting pharmacophores based on Boolean logic, for instance, identifying pharmacophores linked to one kinase but not another. Using the kinase chemogenomic dataset PKIS2 as a starting point, several experimental results predicted were confirmed.