Human Phenotype Project · predictive benchmark
What can each measurement predict?
PhenoBench maps predictive information across 90 clinically grounded tasks, from blood markers and body composition to sleep, glucose dynamics, and longitudinal physiology.
160 matched cells · 52 tasks
Matched model-family leaderboard
Mean rank compares methods only where task, cohort, feature set, split, and metric are fixed. Lower is better. Averaged within each cell, the three pretrained models improved over ridge by a median of 0.004 R² (95% bootstrap CI, 0.002–0.006).
| # | Method | Mean rank | Matched cells |
|---|---|---|---|
| 1 | TabDPT | 2.34 | 160 |
| 2 | TabICL | 2.88 | 160 |
| 3 | TabSwift | 3.17 | 160 |
| 4 | RealMLP | 3.66 | 160 |
| 5 | Ridge | 4.11 | 160 |
| 6 | GBDT | 4.84 | 160 |
Machine-readable evidence: model_family_summary.csv · model_family_contrast.json · model_family_cells.csv
Benchmark atlas
90 clinical tasks across 15 domains
3 framework diagnostics test benchmark mechanics alongside the clinical task set.
Each card links a clinical question to its valid comparison routes. Aggregate result rows contain no participant-level records or private execution identifiers.
- 1-minute orthostatic SBP dropcardiovascular_lipid · vascular_measurement4 tracks · 8 result rows · r2
- Absolute Basophil Countimmune · blood_test4 tracks · 12 result rows · r2
- Absolute Eosinophil Countimmune · blood_test4 tracks · 12 result rows · r2
- Absolute Lymphocyte Countimmune · blood_test4 tracks · 12 result rows · r2
- Absolute Monocyte Countimmune · blood_test4 tracks · 12 result rows · r2
- Absolute Neutrophil Countimmune · blood_test4 tracks · 12 result rows · r2
- Alanine aminotransferase (ALT)hepatobiliary · blood_test2 tracks · 6 result rows · r2
- Albuminhepatobiliary · blood_test2 tracks · 6 result rows · r2
- Apnea–hypopnea index (AHI)respiratory_hypoxaemia · sleep_wearable2 tracks · 6 result rows · r2
- ApoBcardiovascular_lipid · blood_test2 tracks · 4 result rows · r2
- Asthma Statusrespiratory_hypoxaemia · condition_status2 tracks · 6 result rows · roc_auc
- Basophil Percentageimmune · blood_test0 tracks · 0 result rows · r2
- Body Weightanthropometric_body_composition · anthropometry0 tracks · 0 result rows · r2
- Carotid intima-media thicknesscardiovascular_lipid · ultrasound2 tracks · 4 result rows · r2
- CGM coefficient of variationmetabolic_metabolomics · cgm2 tracks · 6 result rows · r2
- CGM mean glucosemetabolic_metabolomics · cgm2 tracks · 6 result rows · r2
- CGM time in range (70–180 mg/dL)metabolic_metabolomics · cgm2 tracks · 6 result rows · r2
- Chronological Ageaging · demographic4 tracks · 6 result rows · r2
- Diastolic blood pressurecardiovascular_lipid · vascular_measurement4 tracks · 8 result rows · r2
- Diet Sleep Causal Effectsleep_physiology · event_level0 tracks · 0 result rows · ate_bpm
- Eosinophil Percentageimmune · blood_test0 tracks · 0 result rows · r2
- Estimated glomerular filtration rate (eGFR)renal_metabolic · derived_clinical_score1 tracks · 3 result rows · r2
- Fasting plasma glucose (FPG)metabolic_metabolomics · blood_test2 tracks · 6 result rows · r2
- Fatty Liver Disease Statushepatobiliary · condition_status2 tracks · 6 result rows · roc_auc
- Fatty liver index (FLI)hepatobiliary · derived_clinical_score2 tracks · 0 result rows · r2
- Femoral neck bone mineral density (BMD)musculoskeletal · dxa_body_composition0 tracks · 0 result rows · r2
- Ferritinhematologic · blood_test2 tracks · 4 result rows · r2
- Fibrosis-3 index (FIB-3)hepatobiliary · derived_clinical_score2 tracks · 6 result rows · r2
- Fibrosis-4 index (FIB-4)hepatobiliary · derived_clinical_score2 tracks · 0 result rows · r2
- Fundus Artery Average Widthophthalmic_retinal · fundus_imaging6 tracks · 18 result rows · r2
- Fundus Artery Fractal Dimensionophthalmic_retinal · fundus_imaging6 tracks · 18 result rows · r2
- Fundus Artery Tortuosity Densityophthalmic_retinal · fundus_imaging6 tracks · 18 result rows · r2
- Fundus Artery Vessel Densityophthalmic_retinal · fundus_imaging6 tracks · 18 result rows · r2
- Fundus Vein Average Widthophthalmic_retinal · fundus_imaging6 tracks · 18 result rows · r2
- Fundus Vein Tortuosity Densityophthalmic_retinal · fundus_imaging6 tracks · 18 result rows · r2
- GlycAmetabolic_metabolomics · nmr_metabolomics2 tracks · 6 result rows · r2
- Glycine (NMR)metabolic_metabolomics · nmr_metabolomics2 tracks · 6 result rows · r2
- Gut Microbiome Identity Retrievalgut_microbiome · microbiome0 tracks · 0 result rows · recall_at_1
- Gut Shannon Diversitygut_microbiome · microbiome2 tracks · 4 result rows · r2
- Gut Species Richnessgut_microbiome · microbiome2 tracks · 4 result rows · r2
- HbA1cmetabolic_metabolomics · blood_test0 tracks · 0 result rows · r2
- HDL cholesterol (NMR)cardiovascular_lipid · nmr_metabolomics4 tracks · 8 result rows · r2
- Hemoglobinhematologic · blood_test2 tracks · 4 result rows · r2
- Hyperlipidemia statuscardiovascular_lipid · condition_status2 tracks · 4 result rows · roc_auc
- Hypertension statuscardiovascular_lipid · condition_status2 tracks · 4 result rows · roc_auc
- Hypothyroidism Statusendocrine · condition_status2 tracks · 4 result rows · roc_auc
- Intrinsic Geometryframework · framework_diagnostic0 tracks · 0 result rows · rankme_normalized
- Irritable Bowel Syndrome Statusgut_microbiome · condition_status2 tracks · 4 result rows · roc_auc
- LDL cholesterolcardiovascular_lipid · blood_test2 tracks · 4 result rows · r2
- LDL cholesterol (NMR)cardiovascular_lipid · nmr_metabolomics2 tracks · 4 result rows · r2
- Lean Body Massanthropometric_body_composition · dxa_body_composition2 tracks · 6 result rows · r2
- Liver Attenuationhepatobiliary · ultrasound2 tracks · 6 result rows · r2
- Liver Stiffnesshepatobiliary · ultrasound2 tracks · 6 result rows · r2
- Loneliness Statusneurologic_psychosocial · questionnaire1 tracks · 3 result rows · balanced_accuracy
- Lumbar spine bone mineral density (BMD)musculoskeletal · dxa_body_composition0 tracks · 0 result rows · r2
- Lymphocyte Percentageimmune · blood_test0 tracks · 0 result rows · r2
- Maximum Bilateral Grip Strengthmusculoskeletal · functional_test2 tracks · 6 result rows · r2
- Meal-conditioned CGM trajectorymetabolic_metabolomics · event_level0 tracks · 0 result rows · rmse
- Mean ankle pressurecardiovascular_lipid · vascular_measurement4 tracks · 8 result rows · r2
- Migraine Statusneurologic_psychosocial · condition_status2 tracks · 6 result rows · roc_auc
- Minimum bilateral ABIcardiovascular_lipid · vascular_measurement4 tracks · 8 result rows · r2
- Minimum nocturnal oxygen saturation (SpO₂)respiratory_hypoxaemia · sleep_wearable2 tracks · 6 result rows · r2
- Monocyte Percentageimmune · blood_test0 tracks · 0 result rows · r2
- Neutrophil Percentageimmune · blood_test0 tracks · 0 result rows · r2
- Next Night Sleepsleep_physiology · event_level3 tracks · 6 result rows · r2
- Non-HDL cholesterolcardiovascular_lipid · blood_test4 tracks · 8 result rows · r2
- Obesity Statusanthropometric_body_composition · condition_status0 tracks · 0 result rows · roc_auc
- Oxygen desaturation index (ODI)respiratory_hypoxaemia · sleep_wearable2 tracks · 6 result rows · r2
- Participant Synthetic Regressionframework · framework_diagnostic0 tracks · 0 result rows · r2
- Postprandial glucose response (PPGR)metabolic_metabolomics · event_level2 tracks · 4 result rows · r2
- PR intervalcardiovascular_lipid · ecg4 tracks · 8 result rows · r2
- Prediabetes Statusmetabolic_metabolomics · condition_status2 tracks · 6 result rows · roc_auc
- QRS durationcardiovascular_lipid · ecg4 tracks · 8 result rows · r2
- QT intervalcardiovascular_lipid · ecg4 tracks · 8 result rows · r2
- QTccardiovascular_lipid · ecg4 tracks · 8 result rows · r2
- Resting heart ratecardiovascular_lipid · ecg2 tracks · 4 result rows · r2
- Sleep Efficiencysleep_physiology · sleep_wearable2 tracks · 4 result rows · r2
- Synthetic Regressionframework · framework_diagnostic0 tracks · 0 result rows · r2
- Systolic blood pressurecardiovascular_lipid · vascular_measurement2 tracks · 4 result rows · r2
- T2D Statusmetabolic_metabolomics · condition_status1 tracks · 3 result rows · roc_auc
- Thigh-ankle pulse-wave velocitycardiovascular_lipid · vascular_measurement2 tracks · 4 result rows · r2
- Thyroid-stimulating hormone (TSH)endocrine · blood_test2 tracks · 4 result rows · r2
- Total branched-chain amino acids (NMR)metabolic_metabolomics · nmr_metabolomics2 tracks · 6 result rows · r2
- Total Sleep Timesleep_physiology · sleep_wearable2 tracks · 4 result rows · r2
- Triglyceridescardiovascular_lipid · blood_test2 tracks · 4 result rows · r2
- Triglycerides (NMR)cardiovascular_lipid · nmr_metabolomics2 tracks · 4 result rows · r2
- Triglyceride–glucose index (TyG)metabolic_metabolomics · derived_clinical_score2 tracks · 6 result rows · r2
- Uraterenal_metabolic · blood_test2 tracks · 6 result rows · r2
- Visceral adipose tissue (VAT)anthropometric_body_composition · dxa_body_composition0 tracks · 0 result rows · r2
- Vitamin Dendocrine · blood_test2 tracks · 4 result rows · r2
- Waistanthropometric_body_composition · anthropometry0 tracks · 0 result rows · r2
- Waist Hip Ratioanthropometric_body_composition · anthropometry0 tracks · 0 result rows · r2
- White blood cell count (WBC)immune · blood_test4 tracks · 12 result rows · r2
Machine-readable catalog: catalog.json
Proof of concept
PhenoBench-LLM
Descriptive validation results on 11 shared tasks. Win rate is the fraction of same-task model comparisons won. It does not measure participant-level prediction accuracy. Models fitted on the same fields generally performed better. The all-available analysis remains downloadable with task coverage.
Machine-readable results: common-task win rates · all-available win rates · task rows
Reproducibility package: Prompts, schemas, synthetic example, and evaluation code.
Maintainer-run evaluation
Submit a container
Develop against fabricated, schema-faithful examples. Then send an OCI image pinned by digest. Maintainers run accepted images on private evaluation data when capacity allows.
- Choose a synthetic task bundle
- Package the prediction contract
- Open a submission issue with the immutable image digest
Source, Task Cards, and contribution guide on GitHub.
Manual intake. No automated queue, turnaround guarantee, or guarantee of publication.