MoleculeMind Shows AI Drug Design Across Two Modalities at WAIC 2026: Nanobodies and Cyclic Peptides Validated Within 30 Days

At the 2026 World Artificial Intelligence Conference (WAIC 2026), AI biotech company MoleculeMind presented new results from its MMDesign platform for de novo cyclic peptide design. The platform designed candidate cyclic peptides targeting protein-protein interaction (PPI) targets including p53-MDM2 and PD-L1, all of which received wet-lab validation.

Just one month earlier, MoleculeMind had published results on de novo nanobody design using the same platform. Two rounds of wet-lab validation in rapid succession signal that the company's AI capability is extending from single-task models toward a cross-modal molecular design platform.

PPI targets are closely tied to oncology and immunology research, but the large and complex protein binding interfaces make traditional small-molecule drug development challenging. Cyclic peptides, with their smaller molecular size and ability to recognize complex targets, have become an important direction in innovative drug development. Traditional approaches rely on large-scale peptide library screening with high early-stage trial-and-error costs.

MoleculeMind's data shows that on the p53-MDM2 target, 16 out of 21 candidate cyclic peptides from MMDesign's first design round received experimental hits, achieving a hit rate exceeding 70%. The best candidate achieved a binding affinity (KD) of 13.7 micromolar, outperforming the experimental positive control. On the PD-L1 target, 6 out of 19 first-round candidates were confirmed to bind the target via surface plasmon resonance (SPR) experiments, with the best KD approaching 40 micromolar.

Compared with global published results on AI-driven peptide design, these first-round hit rates and binding activities place the work at a leading level. The results demonstrate that AI can significantly compress the candidate space before wet-lab experimentation, allowing research teams to prioritize the most promising molecules for experimental resources.

A month earlier, MoleculeMind reported that on over a dozen real therapeutic targets, MMDesign consistently achieved nanomolar to picomolar affinity molecules while designing no more than 50 candidates per target. Nanobodies and cyclic peptides differ substantially in molecular size, structural features, and design constraints. Achieving de novo design validation across both modalities in a short period is a key indicator of platform generalization capability.

MoleculeMind founder Xu Jinbo stated that the key to AI-driven drug development is not single-shot generation or prediction, but building a closed loop of generation, computation, experimentation, and iteration, continuously calibrating models with real experimental data. The company noted that AI4S has moved beyond competition on isolated models and tasks toward competition in systematic scientific discovery capabilities.

The MMDesign capabilities have been integrated into MoleculdOS, the company's self-developed AI macromolecular design operating system. MoleculeMind plans to gradually open access to pharmaceutical companies, universities, research institutes, and innovation teams, unifying models, design tools, data assets, and workflows into a single platform.