{"id":"W4412049114","doi":"10.1016/j.jbi.2025.104873","title":"Accounting for population structure in deep learning models for genomic analysis","year":2025,"lang":"en","type":"article","venue":"Journal of Biomedical Informatics","topic":"Genetic Associations and Epidemiology","field":"Biochemistry, Genetics and Molecular Biology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Alberta Children's Hospital; University of Calgary","funders":"National Institute of Mental Health; Natural Sciences and Engineering Research Council of Canada; National Institutes of Health; Alberta Innovates; Alberta Children's Hospital Foundation; Calgary Foundation","keywords":"Artificial intelligence; Confounding; Deep learning; Population; Computer science; Machine learning; Single-nucleotide polymorphism; Computational biology; Biology; Genotype; Genetics; Statistics; Medicine; Mathematics; Gene","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0007483545,0.00006960873,0.0002564593,0.0002967658,0.00005865494,0.00001560796,0.000118658,0.0001973346,0.000002569034],"category_scores_gemma":[0.0006138732,0.00005836135,0.000172683,0.0002332701,0.00002108385,0.000009494335,0.0000318962,0.0001053578,7.951311e-8],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003763191,"about_ca_system_score_gemma":0.00006520339,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00000597743,"about_ca_topic_score_gemma":0.00002982644,"domain_scores_codex":[0.9989004,0.00002563957,0.0007873861,0.0000510692,0.00007756985,0.0001579645],"domain_scores_gemma":[0.999149,0.00008704076,0.0004977951,0.00006674186,0.0001619913,0.00003742439],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002856146,0.0001053808,0.2128438,0.0003284845,0.001997737,4.378736e-7,0.0008366019,0.6923614,0.01105348,0.0007518891,0.002719342,0.07671583],"study_design_scores_gemma":[0.001501919,0.0002864553,0.06891589,0.00002703997,0.0003844032,0.000004761619,0.0007636228,0.9104089,0.0001549506,0.005604834,0.01181134,0.0001358986],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5273313,0.0001254342,0.4722242,0.0001159618,0.00009044065,0.00008056997,0.000009218075,9.690732e-7,0.00002191396],"genre_scores_gemma":[0.9459961,0.00006950731,0.05332262,0.0002477797,0.0001137069,0.000004243123,0.0002048475,0.000003835879,0.0000373308],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4189016,"threshold_uncertainty_score":0.2379907,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01066905425319228,"score_gpt":0.2878423783043185,"score_spread":0.2771733240511262,"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."}}