{"id":"W4377134072","doi":"10.1101/2023.05.17.541170","title":"Adversarial training improves model interpretability in single-cell RNA-seq analysis","year":2023,"lang":"en","type":"preprint","venue":"bioRxiv (Cold Spring Harbor Laboratory)","topic":"Single-cell and spatial transcriptomics","field":"Biochemistry, Genetics and Molecular Biology","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Princess Margaret Cancer Centre; Lunenfeld-Tanenbaum Research Institute; University Health Network; University of Toronto; Sinai Health System; University of Waterloo","funders":"","keywords":"Interpretability; Robustness (evolution); Machine learning; Artificial intelligence; Computer science; Adversarial system; Deep learning; Training set; Biology; Gene","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003716278,0.001048494,0.0007110355,0.000381551,0.000295771,0.0008284647,0.0009357317,0.001157845,0.001575458],"category_scores_gemma":[0.009435558,0.0003869938,0.000722628,0.0002331405,0.001677425,0.001086247,0.001071568,0.002507895,0.0004195383],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009713647,"about_ca_system_score_gemma":0.0006488625,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002888126,"about_ca_topic_score_gemma":0.002255073,"domain_scores_codex":[0.999255,0.0003878635,0.00002867285,0.0001784119,0.00009041455,0.00005956169],"domain_scores_gemma":[0.9921553,0.006524626,0.0002834804,0.0006586884,0.0002290963,0.00014882],"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":[0.0002113517,0.00004550827,0.002614044,0.00004759366,0.00006500742,0.0000919365,0.00005417595,0.9712405,0.00897833,0.003218085,0.00130554,0.01212791],"study_design_scores_gemma":[0.000002470077,0.000009190325,0.0001855114,0.000003288658,0.000002774636,0.000006603767,0.000003333891,0.9955691,0.001645157,0.002471523,0.00009773944,0.000003401338],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.2928084,0.0006241459,0.6983048,0.002169552,0.0001975729,0.00005796431,0.0007545363,0.002715652,0.002367453],"genre_scores_gemma":[0.9410815,0.0001583653,0.05518994,0.0005804769,0.00005163348,0.00004877054,0.0008789466,0.0002673314,0.00174296],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.003716278,"threshold_uncertainty_score":0.0196538,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02844355033854113,"score_gpt":0.2263122791002056,"score_spread":0.1978687287616645,"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."}}