{"id":"W4411485357","doi":"10.1038/s41746-025-01749-1","title":"Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases","year":2025,"lang":"en","type":"article","venue":"npj Digital Medicine","topic":"Genomics and Rare Diseases","field":"Biochemistry, Genetics and Molecular Biology","cited_by":25,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"Wellcome Trust; National Human Genome Research Institute; National Center for Advancing Translational Sciences; National Institute of Neurological Disorders and Stroke; U.S. Air Force; European Bioinformatics Institute; Harvard Data Science Initiative, Harvard University; Common Fund; Microsoft Research; AstraZeneca; National Institutes of Health; National Science Foundation","keywords":"Disease; Clinical phenotype; Medicine; Rare disease; Medical diagnosis; Machine learning; Phenotype; Artificial intelligence; Bioinformatics; Computer science; Pathology; Gene; Biology; Genetics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00001337502,0.0001234203,0.0001806767,0.00004587893,0.00004770136,0.00001205183,0.0001176794,0.00004497601,0.00001685883],"category_scores_gemma":[0.0002443691,0.00008971681,0.0000593086,0.00007718264,0.000111279,0.000003888268,0.00005699249,0.00002927765,8.597221e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000007353845,"about_ca_system_score_gemma":0.00006094368,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000003971508,"about_ca_topic_score_gemma":0.000004120368,"domain_scores_codex":[0.9993518,0.000008708315,0.0001682278,0.000227768,0.0000983799,0.0001451444],"domain_scores_gemma":[0.9994875,0.00003801576,0.00007153233,0.0001687114,0.0001557702,0.00007848949],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0004032257,0.0002271891,0.9756365,0.0001733211,0.0001349668,0.000002532744,0.00002959471,0.0002426202,0.0006149636,0.00005471227,0.003025692,0.01945465],"study_design_scores_gemma":[0.003719533,0.002303625,0.9391616,0.0003025863,0.0002694606,7.00556e-7,0.0001999561,0.00006327524,0.001767209,0.0002254031,0.0517052,0.0002814771],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.99601,0.001745635,0.000456481,0.0001095819,0.00008892808,0.0003131716,0.0002046153,0.000008771151,0.001062809],"genre_scores_gemma":[0.9981709,0.00009106525,0.00007289807,0.000149802,0.00009769148,0.00006234498,0.001008645,0.00001636089,0.0003303542],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.04867951,"threshold_uncertainty_score":0.3658545,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.005285920508350112,"score_gpt":0.2271593874211226,"score_spread":0.2218734669127725,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}