{"id":"W4407394592","doi":"10.1126/sciadv.adr7338","title":"Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants","year":2025,"lang":"en","type":"article","venue":"Science Advances","topic":"Evolutionary Algorithms and Applications","field":"Computer Science","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Computer science; Benchmark (surveying); Machine learning; Identification (biology); Toolbox; Sampling (signal processing); Artificial intelligence; Data mining; Biology","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":[0.003195775,0.00009783927,0.0001285307,0.0001239715,0.001008154,0.0001541982,0.001529888,0.00002525934,0.000001066604],"category_scores_gemma":[0.0004555622,0.0000716151,0.00003925583,0.001461598,0.0002671895,0.001133027,0.0002003541,0.00009967475,0.000001729273],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004540961,"about_ca_system_score_gemma":0.0002875555,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000263733,"about_ca_topic_score_gemma":0.000001275723,"domain_scores_codex":[0.9985737,0.0001146466,0.0003447388,0.0004881353,0.0002351351,0.0002436626],"domain_scores_gemma":[0.9984965,0.0004351037,0.0002279018,0.0004804412,0.0003258384,0.00003416834],"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":[0.000005047456,0.00004035762,0.000009996018,0.00003052404,0.000004856465,3.380405e-8,0.0001583477,0.005820234,0.6462429,0.1166698,8.364267e-7,0.2310171],"study_design_scores_gemma":[0.0001230394,0.00003332454,0.0004229601,0.0000142811,0.000005912172,5.827306e-7,0.0000564585,0.5723242,0.3813604,0.04504829,0.0005417026,0.00006881093],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.003748037,0.0004514935,0.9932242,0.001274096,0.0002897957,0.0008862071,0.000006129364,0.00008892925,0.00003113196],"genre_scores_gemma":[0.22797,0.00001586271,0.7714032,0.00003890016,0.00001953591,0.0003376076,0.000001208329,0.000002902526,0.0002107829],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.566504,"threshold_uncertainty_score":0.7754011,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03000362555580971,"score_gpt":0.3482759882634606,"score_spread":0.3182723627076509,"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."}}