Modified NSAIDs as potential inhibitors of phosphopeptide recognition by the BRCA1 tandem BRCT domain

Small molecules that disrupt phosphopeptide recognition by the tandem BRCT (tBRCT) domain of BRCA1 are of interest as a means of suppressing homologous recombination (HR) repair and of sensitising tumor cells to poly (ADP-ribose) polymerase inhibitors (PARPi). Here we report an entirely computational design and prioritisation study of amide derivatives assembled from non-steroidal anti-inflammatory drug (NSAID) fragments and intended to occupy the pSer-X-X-Phe recognition cleft of BRCA1 tBRCT. Fragments of ibuprofen, naproxen, flurbiprofen and ketoprofen were combined with chloro- and fluoro-substituted benzylamine or aniline fragments, with and without an intervening methylene linker, to give a library of 64 amides. The library was passed through a sequential in silico workflow: Glide standard-precision (SP) docking into the BRCA1 tBRCT-phosphopeptide structure (PDB 3K0H), Prime MM-GBSA rescoring, induced-fit docking (IFD), binding pose metadynamics (BPMD; ten independent 10 ns replicas per ligand) and triplicate 100 ns molecular dynamics (MD) simulations. Predicted complex structures and binding affinities were additionally obtained with the deep-learning model Boltz-2 across an ensemble of five independent seeds, together with predicted aligned error (PAE) maps. MM-GBSA binding free energies for the library ranged from -30.4 to -53.1 kcal/mol (mean -44.9 ± 3.8). Two compounds were carried forward: 1 (ibuprofen fragment, 2-chlorobenzylamide) and 33 (flurbiprofen fragment, 2-chlorobenzylamide). The static scores and the dynamical assessment rank the library differently: Bractoppin lies seventh of the 65 structures on MM-GBSA and both leads score below it, so the leads were selected on pose stability rather than on score. Comparison with the crystal structure shows that both ligands occupy the phosphopeptide cleft, with the chlorophenyl ring reproducing the position of the conserved Phe4 phenyl ring to within 1.8-2.3 Å. Reporting BPMD as a mean over the ten replicas together with its replica-to-replica spread shows that compound 33 was the only ligand of the 39 BPMD runs performed whose hydrogen-bond network was retained in every replica (per-replica persistence 0.667-0.940; PersScore 0.852 ± 0.090, PoseScore 1.76, CompScore -2.50); for all other ligands at least one replica lost these contacts completely. In triplicate 100 ns MD simulations compounds 1 and 33 maintained hydrogen bonds to Thr1700, Leu1701 and Asn1774 and an intermittent π-cation contact with Arg1835 and halogen contacts with Arg1699. Boltz-2 returned predicted pIC₅₀ values (dimensionless, derived from the model's log10(IC₅₀/μM) output) of 7.83 for compound 1 and 8.47 for compound 33, compared with 7.41 for the reference inhibitor Bractoppin, with interface confidence ipTM of 0.84 and 0.88 and complex pLDDT of 0.87 and 0.89, respectively. These differences are small relative to the reported error of the Boltz-2 affinity module and are therefore used only for qualitative ranking and not as evidence of stronger binding. In a re-docking test the protocol reproduces the crystallographic phosphopeptide pose to a heavy-atom RMSD of 1.79 A, the native configuration ranking second of 128 enumerated states, although the dianionic phosphate group is misplaced by more than 3 A. All data presented are predictive; no experimental measurements were performed. The study identifies two NSAID-derived amides as structurally plausible starting points for the development of BRCA1 tBRCT ligands, whose affinity, selectivity with respect to related BRCT domains and cellular activity remain to be established experimentally.