Mirror-Score benchmarks D-peptide design tools against real-world affinity
Researchers introduced Mirror-Score, a new scoring method for evaluating D-peptide designs in drug discovery. The tool assesses heterochiral D-peptide/L-protein complexes and includes a public benchmark of 31 crystal structures from four target families, 18 with verified affinities. The study found that raw ProteinMPNN negative log-likelihood (NLL) rankings do not correlate with measured binding…
Key points
- Mirror-Score benchmarks 31 crystal structures, 18 with verified affinities across four target families
- Raw ProteinMPNN NLL rankings show poor affinity correlation (Spearman 0.19) and vary by target
- Boltz-2 cofolding confidence achieves 0.90 correlation for viral-entry targets but requires family-specific calibration
The authors emphasize that calibration must be target-family-specific due to limited cross-family transferability. They also outline a prospective design protocol for antimicrobial-resistance targets like Pseudomonas aeruginosa’s LasR and LecB, including mirrored structures and diffusion-model inputs. All code, data, and scripts are openly available on GitHub.
The story so far
2 episodes →- Mirror-Score benchmarks D-peptide design tools against real-world affinitythis story
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researchers test whether AI harnesses specialize or just repeat answers · 1 src
- Study finds synthetic embeddings match text for LLM fine-tuning · 1 src
- Researchers build neurosymbolic router that sends math to exact solvers on Raspberry Pi · 1 src
- OpenAI agents breach Hugging Face infrastructure, study shows alignment testing gaps · 1 src
- Simon Willison shares GPT-6.1 Sol pelicans from OpenAI DevDay 2026 · 1 src
Comments
via GitHub Discussions