{
  "abstract": "Recent benchmarks report that deep-learning-based genetic perturbation models fail to outperform uninformative baselines. Here we introduce a positive control baseline and metric calibration framework, showing—across 14 datasets and 18 metrics—that common benchmarking metrics (for example, mean squared error and Pearson Δ ) are frequently miscalibrated, with reduced sensitivity to measure positive model performance. Under well-calibrated metrics, we find that deep-learning-based genetic perturbation models can outperform uninformative baselines.",
  "abstract_shown": true,
  "affiliations": [
    "Shift Bioscience Ltd., Cambridge, UK. henry@shiftbioscience.com",
    "Shift Bioscience Ltd., Cambridge, UK",
    "Department of Computer Science, University of Toronto, Toronto, Ontario, Canada",
    "Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada",
    "Peter Munk Cardiac Centre, University Health Network, Toronto, Ontario, Canada",
    "Vector Institute, Toronto, Ontario, Canada",
    "School of Clinical Medicine, University of Cambridge, Cambridge, UK"
  ],
  "article_type": null,
  "arxiv": null,
  "author_total": 6,
  "authors": [
    {
      "affs": [
        0
      ],
      "name": "Henry E. Miller",
      "orcid": "0000-0003-3756-3918"
    },
    {
      "affs": [
        1
      ],
      "name": "Gabriel Mejia",
      "orcid": "0000-0003-4382-6390"
    },
    {
      "affs": [
        1
      ],
      "name": "Francis Leblanc",
      "orcid": "0000-0001-7619-3097"
    },
    {
      "affs": [
        1
      ],
      "name": "Brendan Swain",
      "orcid": "0000-0001-9924-9284"
    },
    {
      "affs": [
        2,
        3,
        4,
        5
      ],
      "name": "Bo Wang",
      "orcid": "0000-0002-9620-3413"
    },
    {
      "affs": [
        6,
        1
      ],
      "name": "Lucas Paulo de Lima Camillo",
      "orcid": "0000-0001-8166-0761"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/3qqseendhtdqq.bib",
  "cited_by": 13,
  "doi": "10.1038/s41587-026-03307-w",
  "fulltext_url": null,
  "id": "3qqseendhtdqq",
  "issns": [
    "0733-222X",
    "1087-0156",
    "1546-1696",
    "2331-3684"
  ],
  "landing_url": "https://doi.org/10.1038/s41587-026-03307-w",
  "language": "en",
  "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|Gene Regulatory Network Analysis"
    ]
  ],
  "openalex": "W7215048940",
  "pdf_url": "https://www.nature.com/articles/s41587-026-03307-w.pdf",
  "pmcid": null,
  "pmid": "42823510",
  "published": "2026-10-01",
  "publisher": "Nature Portfolio",
  "ris": "https://research.draconitic.com/paper/3qqseendhtdqq.ris",
  "sources": [
    {
      "first_seen": "2026-11-04T13:21:05Z",
      "native_id": "W7215048940",
      "source": "openalex",
      "url": "https://doi.org/10.1038/s41587-026-03307-w"
    }
  ],
  "subjects": [
    {
      "field": "biochemistry",
      "origin": "openalex:1312|Gene Regulatory Network Analysis",
      "subfield": "molecular-biology"
    }
  ],
  "title": "Deep learning perturbation models can outperform baselines on calibrated metrics",
  "title_text": "Deep learning perturbation models can outperform baselines on calibrated metrics",
  "type": "peer-reviewed",
  "updated_at": "2026-11-04T13:21:05Z",
  "url": "https://research.draconitic.com/paper/3qqseendhtdqq",
  "venue": "Nature Biotechnology",
  "versions": []
}