Discovery of novel peptidomimetics against HSP90-HOP interactions towards improved cancer therapeutics using machine learning strategies.

Background: The HSP90-HOP interaction orchestrates transfer of client proteins from HSP70 to HSP90, promoting their conformational maturation and stabilisation and thereby sustaining oncogenic signalling. The study explores a promising alternative for cancer chemotherapy by targeting this interface rather than the traditional ATP-binding site, which may circumvent the toxicity associated with classical HSP90 inhibitors. Despite its therapeutic relevance, the HSP90-HOP interface remains underexplored, particularly in the context of structure-guided peptidomimetic inhibitors, highlighting a critical gap in strategies to modulate proteostasis in cancer. Aim: This study sought to identify a promising peptidomimetic molecule capable of disrupting the HSP90-HOP interface. Methods: A seven-residue template peptide was engineered from a crucial segment of HOP, with hotspot residues identified through Results: The ML model achieved an accuracy of 0.9055 and an ROC-AUC of 0.9537, indicating strong predictive performance of the model. The lead molecule MMs01053537 demonstrated a favourable binding score of -85.92 kcal/mol, along with a robust stability profile during molecular dynamics simulations. To further assess the consistency of the predicted binding mode across trajectory-derived conformations, ensemble docking and MD-enhanced binding free-energy analysis, along with statistical evaluation were performed. Conclusion: Collectively, these findings position MMs01053537 as a potential candidate for disrupting the HSP90-HOP interaction. However, experimental validation remains essential to confirm its therapeutic potential and support further biological evaluation of the compound.