#25 · Structural bio & protein design

Structural information is becoming residue-resolved, usable output for sequence analysis

The dominant thread is structural information converted into residue-resolved, reusable outputs for sequence-level analysis. PDBe-SIFTS provides residue-level UniProt-PDB mapping via MMseqs2 search, reported 22-36x faster than BLASTP, and StrucPTM extracts modified residues from PDB structures using atom-level rules, annotating relative solvent accessibility and secondary structure. A structural grammar study compares 5,168 canonical and truncated human homodimer pairs from AlphaFold DB, classifying interface conservation, ablation, and emergence. EMERALD-UI exposes stable conformations in alternative alignment space, and a substitution-matrix analysis examines amino acid replacement using predicted structures. Other items apply sequence-based machine learning: a U-Net identifies serpin reactive center loops at about 98% per-sequence accuracy, and NativeReady triages native mass spectrometry suitability from sequence. Across these items, structure is increasingly delivered as a scalable, residue-resolved product of analysis rather than only a benchmark feature.