{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,23]],"date-time":"2025-11-23T08:16:27Z","timestamp":1763885787594,"version":"3.45.0"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"1","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AAAI-SS"],"abstract":"<jats:p>Foundation models (FMs) trained on large-scale single-cell\nRNA-seq (scRNA\u2010seq) data have shown strong performance\nacross various biological tasks. These performances are\noften reported across a large set of test benchmarks across\nall samples. However, the pretraining data of these models\nare often highly imbalanced across disease types, patients'\nconditions, and demographics. For instance, disease samples\nare rarer and more challenging to collect, and the\npretraining sets contain many more healthy cells. Such\nimbalances can hurt performance on underrepresented disease\ncases and the equality of the model outcome. To evaluate\nthis hypothesis, we benchmark off-the-shelf scRNA-seq\nfoundation models for cell-type classification in acute\nmyeloid leukemia (AML), a rare but clinically important\ndisease that represents low-prevalence settings. Here,\nbesides overall performance, we conduct subgroup analysis\nof the outcome across cell types and disease conditions\n(clinical timepoints). Our results suggest that despite\nhigh overall F1 scores in cell-type classification,\nperformance drops in disease conditions and varies across\ncell types. These findings highlight a limitation of\ncurrent scRNA-seq foundation models and motivate more\nbalanced pretraining and failure mode analysis rather than\nan overall performance report.<\/jats:p>","DOI":"10.1609\/aaaiss.v7i1.36931","type":"journal-article","created":{"date-parts":[[2025,11,23]],"date-time":"2025-11-23T08:14:19Z","timestamp":1763885659000},"page":"553-557","source":"Crossref","is-referenced-by-count":0,"title":["From Bias to Breakdown: Benchmarking Failure Mode Analysis\nof Single-cell RNA Sequencing Foundation Models in Acute\nMyeloid Leukemia"],"prefix":"10.1609","volume":"7","author":[{"given":"Amirreza","family":"Naziri","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Arash","family":"Asgari","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Aijun","family":"An","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Eleftherios","family":"Sachlos","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Laleh","family":"Seyyed-Kalantari","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"9382","published-online":{"date-parts":[[2025,11,23]]},"container-title":["Proceedings of the AAAI Symposium Series"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI-SS\/article\/download\/36931\/39069","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI-SS\/article\/download\/36931\/39069","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,23]],"date-time":"2025-11-23T08:14:19Z","timestamp":1763885659000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI-SS\/article\/view\/36931"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,23]]},"references-count":0,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,11,23]]}},"URL":"https:\/\/doi.org\/10.1609\/aaaiss.v7i1.36931","relation":{},"ISSN":["2994-4317"],"issn-type":[{"value":"2994-4317","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,11,23]]}}}