{"id":"W7104034538","doi":"10.48620/92223","title":"Resolving data bias improves generalization in binding affinity prediction.","year":2025,"lang":"en","type":"article","venue":"Open Access CRIS of the University of Bern","topic":"Computational Drug Discovery Methods","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Canadian Bulletin of Medical History","funders":"","keywords":"Benchmark (surveying); Generalization; Training set; Graph; Labeled data; Artificial neural network; Retraining; Test data","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":["open_science"],"consensus_categories":["open_science"],"category_scores_codex":[0.0007877822,0.00006151536,0.0001409439,0.0001937162,0.0001621434,0.0001944117,0.008668323,0.00003201493,0.00001091614],"category_scores_gemma":[0.0002013448,0.00006096581,0.00003392633,0.0009445942,0.00006731789,0.003347983,0.01366332,0.00008054406,7.823302e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006310772,"about_ca_system_score_gemma":0.0002155884,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004080179,"about_ca_topic_score_gemma":0.0005916457,"domain_scores_codex":[0.9990259,0.0002524487,0.0001682779,0.0002789678,0.0001866243,0.00008775653],"domain_scores_gemma":[0.9986084,0.0002239376,0.000207091,0.0008504901,0.00009323432,0.00001691054],"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":[0.0002322299,0.000750697,0.3703685,0.0004252419,0.0002878808,0.00001148041,0.004214484,0.3744692,0.009541818,0.09944274,0.04382648,0.09642928],"study_design_scores_gemma":[0.0007252392,0.00001324727,0.3524554,0.0002698993,0.0000315145,6.506997e-7,0.0004014187,0.6321577,0.005683119,0.006047097,0.002089093,0.000125618],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5547975,0.00005457705,0.4369887,0.002136428,0.0004100726,0.0003348716,0.00007693174,0.00002010595,0.005180798],"genre_scores_gemma":[0.9419196,0.00004011011,0.05696811,0.00007444803,0.000009302462,2.268045e-7,0.00002467301,0.000003775645,0.0009597343],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3871221,"threshold_uncertainty_score":0.9966953,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1290013584496259,"score_gpt":0.376518259563397,"score_spread":0.2475169011137711,"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."}}