From De Novo Molecular Design to Experimentally Validated Protein-Protein Interaction Modulators
Year: 2026
Type: preprint
Abstract: Identifying bioactive small molecules requires balancing target affinity with physicochemical and pharmacokinetic properties compatible with drug development. This is particularly challenging in the development of small-molecule modulators of protein–protein interactions (PPIs) due to the structural complexity of protein interfaces. Here, we developed an integrated workflow combining de novo molecular design with similarity-guided identification of existing compounds from public chemical libraries. We applied this strategy to neuronal calcium sensor-1 (NCS-1), a calcium-binding protein whose interactions with multiple partners regulate synaptic function, making it an attractive target for modulating neuronal signaling. REINVENT4-based molecular generation incorporated predicted binding to the NCS-1 hydrophobic crevice, synthetic accessibility, and CNS-relevant ADMET properties, including blood–brain barrier permeability. Generated structures guided the selection of commercially available analogues. Experimental validation supported the effectiveness of this AI-guided strategy: 11 of the 12 tested compounds bound NCS-1, including 5 with micromolar affinities. In situ protein–protein interaction assays further confirmed the modulatory activity of selected compounds, identifying, to our knowledge, the first small-molecule enhancers of the NCS-1/cannabinoid receptor 1 (CB1R) interaction. CB1R is an attractive pharmacological target implicated in neuronal signaling, whose activity is regulated by NCS-1. High-resolution crystal structures revealed how these modulators engage the NCS-1 hydrophobic crevice, providing structural insights into ligand recognition and suggesting a mechanism of action. These findings connect generative molecular design with experimentally accessible NCS-1 ligands and provide chemical tools and structural insights for investigating NCS-1-dependent regulation of CB1R.
Source: ChemRxiv
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7236464803
