{
  "abstract": "Interface area can dominate discrimination between physiological protein interfaces and crystal contacts. We tested whether local contact organization remains informative after area matching and sequence-cluster separation. In 2,000 balanced BioMany/XtalMany interfaces, sequence cluster aware cross validation gave ROC AUC 0.904 for area and 0.937 after adding five normalized contact descriptors (difference 0.033; bootstrap 95% interval 0.015–0.058). We then fixed 40 biological and 40 crystal homomeric interfaces from the Duarte–Capitani collection, matched pairwise on area within 800–2000 Å2 and sharing no 30%-identity cluster with development. Area reached AUC 0.523; the frozen contact score reached 0.849 (dependency-component bootstrap 95% interval 0.751–0.927) and average precision 0.812. Biological interfaces carried 2.5 times the hydrophobic carbon-pair density of crystal contacts, 1.91 versus 0.76 pairs per 100 Å2. Post hoc analysis identified hydrophobic density as the main contributor; this descriptor alone reached AUC 0.829. On the same panel, unchanged PRODIGY-cryst reached AUC 0.8475 and the lineage-linked current EPPIC score reached 0.876. Contact organization therefore discriminated the curated classes after area matching and sequence-cluster separation. The inference applies to EPPIC-derived labels; protein chain size remained unmatched and associated with the score.",
  "abstract_shown": true,
  "affiliations": [
    "Department of Chemistry and Biochemistry",
    "Stern College for Women, Yeshiva University"
  ],
  "article_type": null,
  "arxiv": null,
  "author_total": 1,
  "authors": [
    {
      "affs": [
        0,
        1
      ],
      "name": "Olga Lavinda",
      "orcid": "0000-0003-3577-5984"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/6ppkn97k4170h.bib",
  "cited_by": 0,
  "doi": "10.26434/chemrxiv.15010687/v1",
  "fulltext_url": null,
  "id": "6ppkn97k4170h",
  "issns": [],
  "landing_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010687/v1",
  "language": null,
  "licence": "CC-BY",
  "licence_label": "CC BY",
  "licence_source": "journal",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "openalex",
      "1312|Protein Structure and Dynamics"
    ]
  ],
  "openalex": "W7223003061",
  "pdf_url": "https://chemrxiv.org/doi/pdf/10.26434/chemrxiv.15010687/v1",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-20",
  "publisher": "American Chemical Society (ACS)",
  "ris": "https://research.draconitic.com/paper/6ppkn97k4170h.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:24Z",
      "native_id": "10.26434/chemrxiv.15010687",
      "source": "chemrxiv",
      "url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010687/v1"
    }
  ],
  "subjects": [
    {
      "field": "biochemistry",
      "origin": "openalex:1312|Protein Structure and Dynamics",
      "subfield": "molecular-biology"
    }
  ],
  "title": "Contact organization separates biological from crystal protein interfaces matched on area and disjoint in sequence",
  "title_text": "Contact organization separates biological from crystal protein interfaces matched on area and disjoint in sequence",
  "type": "preprint",
  "updated_at": "2026-10-30T12:17:24Z",
  "url": "https://research.draconitic.com/paper/6ppkn97k4170h",
  "venue": "ChemRxiv",
  "versions": []
}