Computational de novo design of an α-helical peptide targeting LIMD2 suppresses breast cancer cell proliferation.

Overexpression of LIMD2 (LIM domain-containing protein 2) promotes cancer progression by enhancing cell migration and suppressing apoptosis. Here, we report a computational pipeline for the de novo design of LIMD2-targeting peptide inhibitors. We identified three key binding sites on LIMD2 that form a steric triangle and used RFdiffusion exclusively for geometry-constrained backbone generation within this predefined cavity, a step that does not involve affinity scoring but ensures spatial compatibility prior to sequence design. To enhance structural stability, we constrained sequence design to α-helical peptides using ProteinMPNN. Predicted binding affinities between LIMD2 and designed peptides were evaluated via the deep learning model PEP-PPI, and the top 100 candidates were structurally modeled and refined using Rosetta Build module and FlexPepDock. The top 20 peptides were subjected to toxicity prediction, and finally 17 peptides without toxicity were synthesized and experimentally validated via isothermal titration calorimetry (ITC), with three showing measurable binding, and the best displaying an affinity of 93.7 μM. Top peptides (MSPEERAALAR) exhibited inhibitory effects on tumor cell proliferation in in-vitro assays using triple-negative breast cancer cell lines MDA-MB-231 and MDA-MB-468, highlighting their potential as tumor inhibitors. In further, we analyzed the binding mechanism of the peptide inhibitors against LIMD2, revealing key binding residue within peptide, highlighting the designing binding craft as can identify key sites within peptide to bind with the target. This study presents a robust strategy for de novo functional peptide design and identifies LIMD2-targeting peptides with potential for further antitumor drug development.