{
  "abstract": "Oseltamivir and zanamivir are the primary antivirals against H5N1 influenza, yet resistance profiles at the structural level remain incompletely characterized. Phenotypic screening requires one to two weeks in biosafety level 3 containment; during an active spillover event, this timescale is incompatible with clinical decisions. We present ARGUS (Affinity-and Representation-Guided Unified Scoring), a learned joint model combining four independent structural signals into a unified resistance risk framework: molecular docking affinity, protein language model evolutionary likelihood, perturbation response scanning, and all-atom pocket volume analysis. This work comprises two major components: (1) a comprehensive dual-drug saturation mutagenesis of the H5N1 neuraminidase binding pocket (1,102 mutations, 551 per drug) revealing a striking 166:1 asymmetry between oseltamivir-specific and zanamivir-specific resistance mutations, reflecting zanamivir’s substrate-mimicking pharmacophore; and (2) a novel four-signal learned joint model validated against 37 published fold-resistance measurements spanning H5N1, H1N1pdm09, H3N2, and H7N9 backgrounds (oseltamivir fold-resistance was the fitting target). The four-signal model achieves leave-one-out Spearman ρ = 0.628 (p = 0.000032), with a 95% bootstrap interval of [0.428, 0.845] for the in-sample correlation and a permutation p < 0.001. The earlier two-signal linear blend reached ρ = 0.659 (p = 0.0142) on only n=8 mutations, so the two figures are not directly comparable. The learned weights make docking the dominant signal (96.3% of the total weight) with ESM-2 LLR at 3.7%, which fits the view that protein language models alone conflate evolutionary constraint with drug-resistance specificity. Applied to the 2024 US H5N1 B3.13 outbreak, all 21 outbreak mutations are pocket-distal for both drugs, predicting maintained efficacy of both antivirals. ARGUS provides immediate structural assessment of any novel neuraminidase mutation without experimental work, and the full saturation landscape and trained models are released as a public resource on Zenodo.",
  "abstract_shown": true,
  "affiliations": [
    "School of Science and Technology"
  ],
  "article_type": null,
  "arxiv": null,
  "author_total": 2,
  "authors": [
    {
      "affs": [
        0
      ],
      "name": "Muhammadjon Tursunbadalov",
      "orcid": "0009-0002-6308-1648"
    },
    {
      "affs": [
        0
      ],
      "name": "Mustafojon Tursunbadalov"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/f9ytjd02s0e84.bib",
  "cited_by": 0,
  "doi": "10.26434/chemrxiv.15010305/v1",
  "fulltext_url": null,
  "id": "f9ytjd02s0e84",
  "issns": [],
  "landing_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010305/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",
      "2713|Influenza Virus Research Studies"
    ]
  ],
  "openalex": "W7221617768",
  "pdf_url": "https://chemrxiv.org/doi/pdf/10.26434/chemrxiv.15010305/v1",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-12",
  "publisher": "American Chemical Society (ACS)",
  "ris": "https://research.draconitic.com/paper/f9ytjd02s0e84.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:22Z",
      "native_id": "10.26434/chemrxiv.15010305",
      "source": "chemrxiv",
      "url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010305/v1"
    }
  ],
  "subjects": [
    {
      "field": "medicine",
      "origin": "openalex:2713|Influenza Virus Research Studies",
      "subfield": "epidemiology-public-health"
    }
  ],
  "title": "ARGUS: A Four-Signal Joint Model for H5N1 Neuraminidase Inhibitor Resistance Prediction",
  "title_text": "ARGUS: A Four-Signal Joint Model for H5N1 Neuraminidase Inhibitor Resistance Prediction",
  "type": "preprint",
  "updated_at": "2026-10-30T12:17:22Z",
  "url": "https://research.draconitic.com/paper/f9ytjd02s0e84",
  "venue": "ChemRxiv",
  "versions": []
}