{"id":"W4414598938","doi":"10.1101/2025.09.10.25334730","title":"EAGLE-AI: A large language model workflow for automated extraction and scoring of literature evidence linking genes to autism spectrum disorder","year":2025,"lang":"en","type":"preprint","venue":"medRxiv","topic":"Biomedical Text Mining and Ontologies","field":"Biochemistry, Genetics and Molecular Biology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; SickKids Foundation; Hospital for Sick Children","funders":"University of Toronto; Hospital for Sick Children; Autism Speaks","keywords":"Workflow; Autism; Autism spectrum disorder; Set (abstract data type); Data curation; Component (thermodynamics); Data set; Genomics","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":[0.0003569932,0.0002334829,0.0003075012,0.0001136454,0.00007217211,0.00005518238,0.0002576404,0.0005744972,0.000001837828],"category_scores_gemma":[0.0003317059,0.0002111013,0.0001171683,0.0001266597,0.00004930408,0.000004524642,0.0005380828,0.0002707343,4.596257e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001344958,"about_ca_system_score_gemma":0.00010975,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001744872,"about_ca_topic_score_gemma":0.00005606807,"domain_scores_codex":[0.9986867,0.00005068947,0.0002806385,0.0005823579,0.0001228056,0.0002768285],"domain_scores_gemma":[0.9992799,0.00005929315,0.0001258486,0.0004116489,0.00005069782,0.00007258244],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000925291,0.0003832222,0.06695655,0.007456001,0.0006734423,0.0000417504,0.00476118,0.01459953,0.6891682,0.0004911469,0.00216669,0.212377],"study_design_scores_gemma":[0.002793877,0.001130081,0.03990567,0.02634092,0.0005808083,0.00004879017,0.0004826128,0.6336321,0.2464421,0.004928815,0.04107967,0.002634531],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8191264,0.02981206,0.1485805,0.001335549,0.0004345165,0.0004036146,0.0001689047,0.0001137439,0.00002474137],"genre_scores_gemma":[0.9588794,0.001739111,0.0376993,0.0002021444,0.0001697388,0.0001139824,0.0001632952,0.00002129833,0.001011776],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6190326,"threshold_uncertainty_score":0.8608463,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02099555704033951,"score_gpt":0.3390197854823722,"score_spread":0.3180242284420327,"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."}}