SinCAA model predicts properties of non-canonical amino acid peptides

The wire

1. 01Writing in Advanced Science, the authors report SinCAA, a pretraining framework for non-canonical amino acids that they say beats existing pretrained models on zero-shot peptide property prediction benchmarks.
2. 02Most modified and macrocyclic peptide candidates rely on non-canonical residues, so in silico ranking tools affect which molecules reach the bench.
3. 03Results are computational benchmarks only, with no wet-lab or pharmacokinetic validation reported.

From the source

Similarity-Enhanced Representation Learning of Non-Canonical Amino Acids for Therapeutic Peptide Modeling

Advanced science (Weinheim, Baden-Wurttemberg, Germany) · 2026 Sep 1 · Xu C, Wei L, Wang J et al.