{"id":"W4414084489","doi":"10.1038/s41551-025-01481-x","title":"Combinatorial prediction of therapeutic perturbations using causally inspired neural networks","year":2025,"lang":"en","type":"article","venue":"Nature Biomedical Engineering","topic":"Computational Drug Discovery Methods","field":"Computer Science","cited_by":18,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"Harvard Data Science Initiative, Harvard University; Harvard University; AstraZeneca; Sanofi; Bill and Melinda Gates Foundation; Pfizer; Advanced Research Projects Agency; National Institutes of Health; National Science Foundation","keywords":"Embedding; Representation (politics); Artificial neural network; Graph; Perturbation (astronomy); Drug target; Clinical phenotype; Reversing","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":[0.0002989096,0.0001269734,0.0001753933,0.0002846184,0.00005672724,0.00004329896,0.000433178,0.0002703555,0.000003544002],"category_scores_gemma":[0.000246137,0.0001209947,0.00007358925,0.001192045,0.00004430473,0.0002169945,0.0001552872,0.0004926937,1.830049e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009863199,"about_ca_system_score_gemma":0.0001307825,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000006698128,"about_ca_topic_score_gemma":1.702571e-7,"domain_scores_codex":[0.9988621,0.00005710665,0.0002859224,0.0002405634,0.0003637971,0.0001904794],"domain_scores_gemma":[0.9991429,0.0003926555,0.00005631669,0.0002332783,0.0001046868,0.00007017756],"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":[0.000009769999,0.00008794118,0.0001062098,0.00004643655,0.00008152513,0.000004476788,0.00008469739,0.7105353,0.004452656,0.2676871,0.0000877735,0.0168161],"study_design_scores_gemma":[0.0003491353,0.00003136324,0.005915167,0.00005195561,0.00001523288,0.00000484165,0.00000205891,0.9902471,0.0001249621,0.001965096,0.001214779,0.00007832153],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01713017,0.0007913996,0.9757154,0.0004042853,0.005599914,0.0001181965,0.000007116065,0.0001586342,0.00007486859],"genre_scores_gemma":[0.9775071,0.000003920693,0.02207266,0.0001467625,0.0002327449,0.000004888024,0.00001644035,0.00000806082,0.000007388869],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.960377,"threshold_uncertainty_score":0.493402,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01050557665194205,"score_gpt":0.2734299059163521,"score_spread":0.2629243292644101,"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."}}