Models / protein-ml / esm2-t36-3b-ur50d / benchmarks

protein-ml/esm2-t36-3b-ur50d

Visibility: PUBLIC

Model card FilesBenchmarks

Group by

Domain variant fitness, human PLM 0.5 (ProteinGym x Pfam) protein-ml/protein-domain-variant-fitness-pfam-05

Drawn from: ProteinGym, ProteinGym_v1

atlas-n25-v2

Metric

Value

95% CI

Rank

Best in catalogue

Source

Overlap

Spearman ρ (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.292 [−0.176, 0.700]	25
MET_HUMAN_Estevam_2023	0.262 [−0.147, 0.642]	25
S22A1_HUMAN_Yee_2023_activity	0.575 [0.305, 0.776]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.348 [0.003, 0.600]	25

0.369

[0.159, 0.522]

100

3 / 13

0.476

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

AUC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.647 [0.413, 0.857]	25
MET_HUMAN_Estevam_2023	0.688 [0.458, 0.897]	25
S22A1_HUMAN_Yee_2023_activity	0.833 [0.667, 0.986]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.722 [0.481, 0.910]	25

0.723

[0.611, 0.816]

100

3 / 13

0.759

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

MCC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.199 [−0.294, 0.603]	25
MET_HUMAN_Estevam_2023	0.277 [−0.131, 0.690]	25
S22A1_HUMAN_Yee_2023_activity	0.621 [0.196, 0.852]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.280 [−0.231, 0.704]	25

0.344

[0.088, 0.526]

100

2 / 13

0.426

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

NDCG, top 10 % (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.247 [0.000, 0.773]	25
MET_HUMAN_Estevam_2023	0.614 [0.387, 0.894]	25
S22A1_HUMAN_Yee_2023_activity	0.898 [0.583, 1.000]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.357 [0.094, 0.836]	25

0.529

[0.374, 0.725]

100

—

0.635

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Top-10 % recall (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.000 [0.000, 0.000]	25
MET_HUMAN_Estevam_2023	0.000 [0.000, 0.338]	25
S22A1_HUMAN_Yee_2023_activity	0.333 [0.000, 1.000]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.000 [0.000, 0.750]	25

0.083

[0.000, 0.375]

100

—

0.167

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Domain variant fitness, human PLM 0.9 (ProteinGym x Pfam) protein-ml/protein-domain-variant-fitness-pfam-09

Drawn from: ProteinGym, ProteinGym_v1

atlas-n25-v2

Metric

Value

95% CI

Rank

Best in catalogue

Source

Overlap

Spearman ρ (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.292 [−0.176, 0.700]	25
MET_HUMAN_Estevam_2023	0.262 [−0.147, 0.642]	25
S22A1_HUMAN_Yee_2023_activity	0.575 [0.305, 0.776]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.348 [0.003, 0.600]	25

0.369

[0.159, 0.522]

100

3 / 13

0.476

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

AUC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.647 [0.413, 0.857]	25
MET_HUMAN_Estevam_2023	0.688 [0.458, 0.897]	25
S22A1_HUMAN_Yee_2023_activity	0.833 [0.667, 0.986]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.722 [0.481, 0.910]	25

0.723

[0.611, 0.816]

100

3 / 13

0.759

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

MCC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.199 [−0.294, 0.603]	25
MET_HUMAN_Estevam_2023	0.277 [−0.131, 0.690]	25
S22A1_HUMAN_Yee_2023_activity	0.621 [0.196, 0.852]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.280 [−0.231, 0.704]	25

0.344

[0.088, 0.526]

100

2 / 13

0.426

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

NDCG, top 10 % (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.247 [0.000, 0.773]	25
MET_HUMAN_Estevam_2023	0.614 [0.387, 0.894]	25
S22A1_HUMAN_Yee_2023_activity	0.898 [0.583, 1.000]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.357 [0.094, 0.836]	25

0.529

[0.374, 0.725]

100

—

0.635

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Top-10 % recall (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
CCR5_HUMAN_Gill_2023	0.000 [0.000, 0.000]	25
MET_HUMAN_Estevam_2023	0.000 [0.000, 0.338]	25
S22A1_HUMAN_Yee_2023_activity	0.333 [0.000, 1.000]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.000 [0.000, 0.750]	25

0.083

[0.000, 0.375]

100

—

0.167

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Domain variant fitness, human PLM Clans-0.5 (ProteinGym x Pfam) protein-ml/protein-domain-variant-fitness-clan-05

Drawn from: ProteinGym, ProteinGym_v1

atlas-n25-v2

Metric

Value

95% CI

Rank

Best in catalogue

Source

Overlap

Spearman ρ (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
ADRB2_HUMAN_Jones_2020	0.207 [−0.199, 0.567]	25
CCR5_HUMAN_Gill_2023	0.292 [−0.176, 0.700]	25
OPSD_HUMAN_Wan_2019	0.278 [−0.171, 0.658]	25
S22A1_HUMAN_Yee_2023_activity	0.575 [0.305, 0.776]	25

0.338

[0.122, 0.518]

100

6 / 13

0.463

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

AUC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
ADRB2_HUMAN_Jones_2020	0.597 [0.350, 0.820]	25
CCR5_HUMAN_Gill_2023	0.647 [0.413, 0.857]	25
OPSD_HUMAN_Wan_2019	0.694 [0.427, 0.912]	25
S22A1_HUMAN_Yee_2023_activity	0.833 [0.667, 0.986]	25

0.693

[0.564, 0.800]

100

4 / 13

0.751

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

MCC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
ADRB2_HUMAN_Jones_2020	0.116 [−0.343, 0.519]	25
CCR5_HUMAN_Gill_2023	0.199 [−0.294, 0.603]	25
OPSD_HUMAN_Wan_2019	0.113 [−0.243, 0.621]	25
S22A1_HUMAN_Yee_2023_activity	0.621 [0.196, 0.852]	25

0.262

[0.014, 0.464]

100

5 / 13

0.426

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

NDCG, top 10 % (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
ADRB2_HUMAN_Jones_2020	0.544 [0.319, 0.912]	25
CCR5_HUMAN_Gill_2023	0.247 [0.000, 0.773]	25
OPSD_HUMAN_Wan_2019	0.833 [0.436, 1.000]	25
S22A1_HUMAN_Yee_2023_activity	0.898 [0.583, 1.000]	25

0.630

[0.473, 0.805]

100

—

0.709

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Top-10 % recall (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
ADRB2_HUMAN_Jones_2020	0.000 [0.000, 0.750]	25
CCR5_HUMAN_Gill_2023	0.000 [0.000, 0.000]	25
OPSD_HUMAN_Wan_2019	0.333 [0.000, 1.000]	25
S22A1_HUMAN_Yee_2023_activity	0.333 [0.000, 1.000]	25

0.167

[0.000, 0.500]

100

—

0.250

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Domain variant fitness, human PLM Clans-0.9 (ProteinGym x Pfam) protein-ml/protein-domain-variant-fitness-clan-09

Drawn from: ProteinGym, ProteinGym_v1

atlas-n25-v2

Metric

Value

95% CI

Rank

Best in catalogue

Source

Overlap

Spearman ρ (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
MET_HUMAN_Estevam_2023	0.262 [−0.147, 0.642]	25
MK01_HUMAN_Brenan_2016	0.428 [−0.005, 0.752]	25
SCN5A_HUMAN_Glazer_2019	−0.022 [−0.519, 0.414]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.348 [0.003, 0.600]	25

0.254

[0.038, 0.435]

100

7 / 13

0.362

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

AUC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
MET_HUMAN_Estevam_2023	0.688 [0.458, 0.897]	25
MK01_HUMAN_Brenan_2016	0.860 [0.581, 1.000]	25
SCN5A_HUMAN_Glazer_2019	0.447 [0.206, 0.686]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.722 [0.481, 0.910]	25

0.679

[0.559, 0.790]

100

4 / 13

0.722

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

MCC (zero-shot, domain input) ↑ higher is better · zero-shot

By assay: 4 levels

assay	Value [95% CI]	n
MET_HUMAN_Estevam_2023	0.277 [−0.131, 0.690]	25
MK01_HUMAN_Brenan_2016	0.487 [0.045, 0.784]	25
SCN5A_HUMAN_Glazer_2019	−0.131 [−0.487, 0.333]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.280 [−0.231, 0.704]	25

0.229

[0.015, 0.438]

100

6 / 13

0.310

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

NDCG, top 10 % (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
MET_HUMAN_Estevam_2023	0.614 [0.387, 0.894]	25
MK01_HUMAN_Brenan_2016	0.399 [0.000, 0.925]	25
SCN5A_HUMAN_Glazer_2019	0.489 [0.000, 1.000]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.357 [0.094, 0.836]	25

0.465

[0.254, 0.747]

100

—

0.725

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Top-10 % recall (zero-shot, domain input) ↑ higher is better · zero-shot · secondary

By assay: 4 levels

assay	Value [95% CI]	n
MET_HUMAN_Estevam_2023	0.000 [0.000, 0.338]	25
MK01_HUMAN_Brenan_2016	0.000 [0.000, 0.750]	25
SCN5A_HUMAN_Glazer_2019	0.333 [0.000, 1.000]	25
SRC_HUMAN_Chakraborty_2023_binding-DAS_25uM	0.000 [0.000, 0.750]	25

0.083

[0.000, 0.429]

100

—

0.250

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

GO function prediction (Swiss-Prot 2025H2) protein-ml/protein-function-go-prediction

Drawn from: UniProt Knowledgebase

atlas-n100-v1

Metric	Value	95% CI	n	Rank	Best in catalogue	Source	Overlap	Board
Fmax (molecular function, probe) ↑ higher is better · probe	0.645	[0.597, 0.728]	100	14 / 17	0.844	reproduced	unknown	Board
Fmax (biological process, probe) ↑ higher is better · probe	0.412	[0.338, 0.538]	100	16 / 17	0.538	reproduced	unknown	Board
Fmax (cellular component, probe) ↑ higher is better · probe	0.756 best	[0.705, 0.812]	100	1 / 17	0.756	reproduced	unknown	Board
Micro AUPR (molecular function, probe) ↑ higher is better · probe · secondary	0.652	[0.576, 0.754]	100	—	0.875	reproduced	unknown	Board
Smin (molecular function, probe) ↓ lower is better · bits · probe · secondary	3.361	[2.526, 4.063]	100	—	1.727	reproduced	unknown	Board
Micro AUPR (biological process, probe) ↑ higher is better · probe · secondary	0.380	[0.283, 0.498]	100	—	0.595	reproduced	unknown	Board
Smin (biological process, probe) ↓ lower is better · bits · probe · secondary	6.120	[4.351, 7.711]	100	—	4.842	reproduced	unknown	Board
Micro AUPR (cellular component, probe) ↑ higher is better · probe · secondary	0.722	[0.656, 0.791]	100	—	0.758	reproduced	unknown	Board
Smin (cellular component, probe) ↓ lower is better · bits · probe · secondary	3.092 best	[2.384, 3.767]	100	—	3.092	reproduced	unknown	Board

Protein membrane association (DeepLoc) protein-ml/protein-membrane-localization

atlas-full-v1

Metric	Value	95% CI	n	Rank	Best in catalogue	Source	Overlap	Board
Q2 accuracy (logistic probe) ↑ higher is better · probe	0.848	[0.831, 0.866]	1728	8 / 17	0.888	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board
MCC (logistic probe) ↑ higher is better · probe · secondary	0.689	[0.653, 0.727]	1728	—	0.771	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board

Protein stability ΔΔG (Mega-scale) protein-ml/protein-stability-ddg

atlas-n100-v1

Metric	Value	95% CI	n	Rank	Best in catalogue	Source	Overlap	Board
Spearman ρ (zero-shot) ↑ higher is better · zero-shot	0.410 best	[0.203, 0.597]	100	1 / 12	0.410	reproduced	unknown	Board
Spearman ρ (ridge probe) ↑ higher is better · probe	0.126	[−0.087, 0.304]	100	9 / 17	0.297	reproduced	unknown	Board
RMSE, kcal/mol (ridge probe) ↓ lower is better · kcal/mol · probe	1.114	[0.888, 1.318]	100	9 / 17	1.095	reproduced	unknown	Board
Pearson r (ridge probe) ↑ higher is better · probe · secondary	0.148	[−0.050, 0.304]	100	—	0.349	reproduced	unknown	Board
MAE, kcal/mol (ridge probe) ↓ lower is better · kcal/mol · probe · secondary	0.833	[0.683, 0.989]	100	—	0.819	reproduced	unknown	Board

Protein subcellular localization (DeepLoc) protein-ml/protein-subcellular-localization

atlas-full-v1

Metric	Value	95% CI	n	Rank	Best in catalogue	Source	Overlap	Board
Accuracy (logistic probe) ↑ higher is better · probe	0.741	[0.727, 0.754]	2742	5 / 17	0.801	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board
Macro F1 (logistic probe) ↑ higher is better · probe	0.544	[0.525, 0.562]	2742	5 / 17	0.640	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board
MCC (logistic probe) ↑ higher is better · probe · secondary	0.685	[0.668, 0.701]	2742	—	0.758	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board

ProteinGym DMS zero-shot (Spearman) protein-ml/proteingym-dms-zero-shot-spearman

Drawn from: ProteinGym, ProteinGym_v1

atlas-n100-v1

Metric

Value

95% CI

Rank

Best in catalogue

Source

Overlap

Spearman ρ (zero-shot) ↑ higher is better · zero-shot

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.603 [0.473, 0.725]	100
CASP3_HUMAN_Roychowdhury_2020	0.605 [0.470, 0.717]	100
CCR5_HUMAN_Gill_2023	0.307 [0.114, 0.496]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.630 [0.475, 0.731]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.235 [0.054, 0.412]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.077 [−0.125, 0.278]	100
MLAC_ECOLI_MacRae_2023	0.283 [0.097, 0.459]	100
MTH3_HAEAE_RockahShmuel_2015	0.584 [0.427, 0.713]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.601 [0.448, 0.725]	100
SPG1_STRSG_Olson_2014	0.092 [−0.127, 0.316]	100

0.402

[0.345, 0.447]

1000

7 / 11

0.434

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

AUC (zero-shot) ↑ higher is better · zero-shot

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.773 [0.681, 0.862]	100
CASP3_HUMAN_Roychowdhury_2020	0.865 [0.791, 0.931]	100
CCR5_HUMAN_Gill_2023	0.667 [0.558, 0.771]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.815 [0.719, 0.886]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.679 [0.576, 0.783]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.673 [0.553, 0.788]	100
MLAC_ECOLI_MacRae_2023	0.698 [0.596, 0.794]	100
MTH3_HAEAE_RockahShmuel_2015	0.851 [0.753, 0.933]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.853 [0.775, 0.924]	100
SPG1_STRSG_Olson_2014	0.817 [0.712, 0.911]	100

0.769

[0.738, 0.798]

1000

2 / 11

0.775

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

MCC (zero-shot) ↑ higher is better · zero-shot

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.380 [0.204, 0.590]	100
CASP3_HUMAN_Roychowdhury_2020	0.541 [0.366, 0.700]	100
CCR5_HUMAN_Gill_2023	0.200 [−0.003, 0.387]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.482 [0.279, 0.642]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.175 [0.022, 0.401]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.312 [0.080, 0.489]	100
MLAC_ECOLI_MacRae_2023	0.320 [0.080, 0.460]	100
MTH3_HAEAE_RockahShmuel_2015	0.375 [0.172, 0.524]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.572 [0.343, 0.721]	100
SPG1_STRSG_Olson_2014	0.346 [0.160, 0.453]	100

0.370 best

[0.293, 0.413]

1000

1 / 11

0.370

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

NDCG, top 10 % (zero-shot) ↑ higher is better · zero-shot

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.795 [0.694, 0.922]	100
CASP3_HUMAN_Roychowdhury_2020	0.804 [0.722, 0.885]	100
CCR5_HUMAN_Gill_2023	0.796 [0.569, 0.887]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.840 [0.767, 0.913]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.825 [0.748, 0.904]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.561 [0.419, 0.673]	100
MLAC_ECOLI_MacRae_2023	0.843 [0.742, 0.917]	100
MTH3_HAEAE_RockahShmuel_2015	0.718 [0.545, 0.836]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.887 [0.802, 0.953]	100
SPG1_STRSG_Olson_2014	0.837 [0.637, 0.938]	100

0.791

[0.741, 0.820]

1000

6 / 11

0.814

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Top-10 % recall (zero-shot) ↑ higher is better · zero-shot

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.100 [0.000, 0.417]	100
CASP3_HUMAN_Roychowdhury_2020	0.200 [0.000, 0.400]	100
CCR5_HUMAN_Gill_2023	0.000 [0.000, 0.300]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.300 [0.077, 0.600]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.000 [0.000, 0.100]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.100 [0.000, 0.300]	100
MLAC_ECOLI_MacRae_2023	0.100 [0.000, 0.333]	100
MTH3_HAEAE_RockahShmuel_2015	0.091 [0.000, 0.417]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.400 [0.083, 0.700]	100
SPG1_STRSG_Olson_2014	0.300 [0.000, 0.500]	100

0.159

[0.078, 0.232]

1000

8 / 11

0.196

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Spearman ρ (ridge probe) ↑ higher is better · probe

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.627 [0.464, 0.770]	100
CASP3_HUMAN_Roychowdhury_2020	0.501 [0.323, 0.644]	100
CCR5_HUMAN_Gill_2023	0.075 [−0.153, 0.300]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.736 [0.618, 0.814]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.746 [0.603, 0.844]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.553 [0.381, 0.691]	100
MLAC_ECOLI_MacRae_2023	0.486 [0.311, 0.642]	100
MTH3_HAEAE_RockahShmuel_2015	0.617 [0.480, 0.721]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.868 [0.807, 0.914]	100
SPG1_STRSG_Olson_2014	0.787 [0.666, 0.870]	100

0.600

[0.551, 0.640]

1000

7 / 17

0.661

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

AUC (ridge probe) ↑ higher is better · probe

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.789 [0.689, 0.878]	100
CASP3_HUMAN_Roychowdhury_2020	0.773 [0.672, 0.858]	100
CCR5_HUMAN_Gill_2023	0.575 [0.453, 0.692]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.855 [0.770, 0.924]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.920 [0.840, 0.979]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.837 [0.739, 0.921]	100
MLAC_ECOLI_MacRae_2023	0.784 [0.695, 0.866]	100
MTH3_HAEAE_RockahShmuel_2015	0.816 [0.716, 0.908]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.959 [0.918, 0.994]	100
SPG1_STRSG_Olson_2014	0.958 [0.918, 0.987]	100

0.827

[0.801, 0.852]

1000

5 / 17

0.848

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

MCC (ridge probe) ↑ higher is better · probe

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.500 [0.301, 0.680]	100
CASP3_HUMAN_Roychowdhury_2020	0.381 [0.161, 0.545]	100
CCR5_HUMAN_Gill_2023	0.160 [−0.040, 0.361]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.602 [0.400, 0.730]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.524 [0.365, 0.684]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.445 [0.241, 0.600]	100
MLAC_ECOLI_MacRae_2023	0.400 [0.241, 0.620]	100
MTH3_HAEAE_RockahShmuel_2015	0.421 [0.251, 0.562]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.694 [0.570, 0.847]	100
SPG1_STRSG_Olson_2014	0.403 [0.289, 0.516]	100

0.453

[0.394, 0.502]

1000

5 / 17

0.472

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

NDCG, top 10 % (ridge probe) ↑ higher is better · probe

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.783 [0.646, 0.933]	100
CASP3_HUMAN_Roychowdhury_2020	0.870 [0.785, 0.941]	100
CCR5_HUMAN_Gill_2023	0.648 [0.363, 0.749]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.853 [0.764, 0.917]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.931 [0.897, 0.965]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.717 [0.590, 0.865]	100
MLAC_ECOLI_MacRae_2023	0.905 [0.841, 0.955]	100
MTH3_HAEAE_RockahShmuel_2015	0.799 [0.634, 0.912]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.891 [0.847, 0.966]	100
SPG1_STRSG_Olson_2014	0.935 [0.887, 0.973]	100

0.833

[0.793, 0.865]

1000

6 / 17

0.860

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Top-10 % recall (ridge probe) ↑ higher is better · probe

By assay: 10 levels

assay	Value [95% CI]	n
ANCSZ_Hobbs_2022	0.200 [0.000, 0.500]	100
CASP3_HUMAN_Roychowdhury_2020	0.300 [0.100, 0.700]	100
CCR5_HUMAN_Gill_2023	0.000 [0.000, 0.273]	100
CP2C9_HUMAN_Amorosi_2021_abundance	0.300 [0.000, 0.636]	100
ILF3_HUMAN_Tsuboyama_2023_2L33	0.400 [0.091, 0.700]	100
KCNE1_HUMAN_Muhammad_2023_expression	0.400 [0.091, 0.667]	100
MLAC_ECOLI_MacRae_2023	0.300 [0.000, 0.545]	100
MTH3_HAEAE_RockahShmuel_2015	0.273 [0.000, 0.600]	100
RPC1_BP434_Tsuboyama_2023_1R69	0.500 [0.200, 0.800]	100
SPG1_STRSG_Olson_2014	0.200 [0.000, 0.545]	100

0.287

[0.196, 0.381]

1000

6 / 17

0.360

reproduced

overlap

Train/test overlap declared by the evaluator: protein-ml/uniref50

Board

Secondary structure (Q3) protein-ml/secondary-structure-q3

Drawn from: tape

atlas-n100-v2

Metric	Value	95% CI	n	Rank	Best in catalogue	Source	Overlap	Board
Q3 accuracy (per-residue probe) ↑ higher is better · probe	0.779	[0.765, 0.791]	100	5 / 17	0.813	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board
Q8 accuracy (per-residue probe) ↑ higher is better · probe	0.629	[0.610, 0.648]	100	5 / 17	0.663	reproduced	overlap Train/test overlap declared by the evaluator: protein-ml/uniref50	Board