MitoInteract / recovery /reports /bindingdb-v2-descriptor-baselines.json
Ethan Troy
Cursor
feat: add source-aware exact-Kd model evaluation with honest no-publish verdict
f9c6388
Raw History Blame Contribute Delete
52.3 kB
{
"sample_rows": 2327,
"target": "pKd",
"seed": 42,
"packages": {
"lightgbm": "4.7.0",
"numpy": "2.5.1",
"rdkit": "2026.3.4"
},
"feature_dimensions": {
"ligand": 2056,
"protein": 421,
"combined": 2477
},
"splits": {
"cold_both": {
"rows": {
"train": 1822,
"validation": 292,
"test": 213
},
"mean_baseline": {
"prediction": 6.300028370334318,
"validation": {
"rmse": 1.7964309076980576,
"mae": 1.4819881649030124,
"pearson_r": null,
"spearman_r": null,
"r2": -0.8086919248505475
},
"test": {
"rmse": 1.8152406225701505,
"mae": 1.5873559607915237,
"pearson_r": null,
"spearman_r": null,
"r2": -1.552108571424582
}
},
"models": [
{
"name": "ligand_morgan_descriptors_lightgbm",
"best_iteration": 167,
"validation": {
"rmse": 1.623502038358803,
"mae": 1.3354019246431537,
"pearson_r": 0.3845765979676542,
"spearman_r": 0.3758109179620909,
"r2": -0.4772338103515368
},
"test": {
"rmse": 2.0432584219380625,
"mae": 1.68112188603478,
"pearson_r": -0.2021519211052322,
"spearman_r": -0.25884521276274436,
"r2": -2.233533299135532
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 380
},
{
"feature": "tpsa",
"gain_proxy": 253
},
{
"feature": "mol_logp",
"gain_proxy": 228
},
{
"feature": "fraction_csp3",
"gain_proxy": 204
},
{
"feature": "rotatable_bonds",
"gain_proxy": 119
},
{
"feature": "h_acceptors",
"gain_proxy": 110
},
{
"feature": "ring_count",
"gain_proxy": 90
},
{
"feature": "morgan_1607",
"gain_proxy": 72
},
{
"feature": "h_donors",
"gain_proxy": 67
},
{
"feature": "morgan_192",
"gain_proxy": 38
},
{
"feature": "morgan_807",
"gain_proxy": 35
},
{
"feature": "morgan_516",
"gain_proxy": 34
},
{
"feature": "morgan_1679",
"gain_proxy": 34
},
{
"feature": "morgan_1956",
"gain_proxy": 34
},
{
"feature": "morgan_637",
"gain_proxy": 33
},
{
"feature": "morgan_1126",
"gain_proxy": 33
},
{
"feature": "morgan_1917",
"gain_proxy": 33
},
{
"feature": "morgan_314",
"gain_proxy": 32
},
{
"feature": "morgan_387",
"gain_proxy": 32
},
{
"feature": "morgan_953",
"gain_proxy": 32
}
],
"fit_and_eval_seconds": 0.6473237399986829
},
{
"name": "protein_aac_dipeptide_lightgbm",
"best_iteration": 20,
"validation": {
"rmse": 1.7271934793224868,
"mae": 1.4443852354500681,
"pearson_r": 0.08175660644771442,
"spearman_r": 0.066573786691369,
"r2": -0.6719586868371021
},
"test": {
"rmse": 1.7888131932137514,
"mae": 1.5874690666318834,
"pearson_r": 0.22255723821723375,
"spearman_r": 0.2404112337040462,
"r2": -1.4783390560870262
},
"top_feature_importance": [
{
"feature": "aac_F",
"gain_proxy": 19
},
{
"feature": "dipeptide_RN",
"gain_proxy": 11
},
{
"feature": "dipeptide_ST",
"gain_proxy": 11
},
{
"feature": "aac_C",
"gain_proxy": 10
},
{
"feature": "dipeptide_II",
"gain_proxy": 10
},
{
"feature": "log_protein_length",
"gain_proxy": 9
},
{
"feature": "dipeptide_CV",
"gain_proxy": 9
},
{
"feature": "dipeptide_SK",
"gain_proxy": 9
},
{
"feature": "dipeptide_AK",
"gain_proxy": 8
},
{
"feature": "dipeptide_MF",
"gain_proxy": 8
},
{
"feature": "dipeptide_TE",
"gain_proxy": 8
},
{
"feature": "dipeptide_LP",
"gain_proxy": 7
},
{
"feature": "dipeptide_LS",
"gain_proxy": 7
},
{
"feature": "dipeptide_PQ",
"gain_proxy": 7
},
{
"feature": "dipeptide_DL",
"gain_proxy": 6
},
{
"feature": "dipeptide_GL",
"gain_proxy": 6
},
{
"feature": "dipeptide_GV",
"gain_proxy": 6
},
{
"feature": "dipeptide_MP",
"gain_proxy": 6
},
{
"feature": "dipeptide_QL",
"gain_proxy": 6
},
{
"feature": "dipeptide_VE",
"gain_proxy": 6
}
],
"fit_and_eval_seconds": 0.41759571899456205
},
{
"name": "combined_morgan_protein_lightgbm",
"best_iteration": 148,
"validation": {
"rmse": 1.5984113304646947,
"mae": 1.3252363203312023,
"pearson_r": 0.17118197154497686,
"spearman_r": 0.14869883824937283,
"r2": -0.4319262856518138
},
"test": {
"rmse": 1.8863991862426097,
"mae": 1.666751331561644,
"pearson_r": 0.28954024607641626,
"spearman_r": 0.3596437278767927,
"r2": -1.7561188612102443
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 245
},
{
"feature": "tpsa",
"gain_proxy": 125
},
{
"feature": "mol_logp",
"gain_proxy": 116
},
{
"feature": "dipeptide_GL",
"gain_proxy": 55
},
{
"feature": "ring_count",
"gain_proxy": 48
},
{
"feature": "fraction_csp3",
"gain_proxy": 45
},
{
"feature": "morgan_347",
"gain_proxy": 40
},
{
"feature": "morgan_1985",
"gain_proxy": 37
},
{
"feature": "dipeptide_II",
"gain_proxy": 35
},
{
"feature": "h_donors",
"gain_proxy": 35
},
{
"feature": "morgan_1766",
"gain_proxy": 34
},
{
"feature": "rotatable_bonds",
"gain_proxy": 34
},
{
"feature": "morgan_1349",
"gain_proxy": 33
},
{
"feature": "h_acceptors",
"gain_proxy": 33
},
{
"feature": "dipeptide_RN",
"gain_proxy": 32
},
{
"feature": "dipeptide_QS",
"gain_proxy": 31
},
{
"feature": "aac_F",
"gain_proxy": 29
},
{
"feature": "morgan_675",
"gain_proxy": 28
},
{
"feature": "morgan_1129",
"gain_proxy": 28
},
{
"feature": "dipeptide_DT",
"gain_proxy": 27
}
],
"fit_and_eval_seconds": 2.515122621000046
}
]
},
"cold_protein_exact": {
"rows": {
"train": 1820,
"validation": 237,
"test": 270
},
"mean_baseline": {
"prediction": 6.29383344109891,
"validation": {
"rmse": 1.660535937737737,
"mae": 1.284219476481831,
"pearson_r": null,
"spearman_r": null,
"r2": -0.00032164778640297875
},
"test": {
"rmse": 1.7658507036994262,
"mae": 1.4882390526038425,
"pearson_r": null,
"spearman_r": null,
"r2": -0.014287223246756575
}
},
"models": [
{
"name": "ligand_morgan_descriptors_lightgbm",
"best_iteration": 77,
"validation": {
"rmse": 1.2123700914112552,
"mae": 0.9397933284640939,
"pearson_r": 0.6902623865127632,
"spearman_r": 0.5667836134543494,
"r2": 0.46677127463597157
},
"test": {
"rmse": 1.1702173443155717,
"mae": 0.960761845204874,
"pearson_r": 0.7660156801203487,
"spearman_r": 0.7458266499379779,
"r2": 0.5545632440824192
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 151
},
{
"feature": "fraction_csp3",
"gain_proxy": 114
},
{
"feature": "tpsa",
"gain_proxy": 90
},
{
"feature": "mol_logp",
"gain_proxy": 88
},
{
"feature": "rotatable_bonds",
"gain_proxy": 67
},
{
"feature": "morgan_1607",
"gain_proxy": 55
},
{
"feature": "ring_count",
"gain_proxy": 51
},
{
"feature": "h_acceptors",
"gain_proxy": 41
},
{
"feature": "morgan_1452",
"gain_proxy": 35
},
{
"feature": "morgan_1745",
"gain_proxy": 33
},
{
"feature": "morgan_1126",
"gain_proxy": 32
},
{
"feature": "h_donors",
"gain_proxy": 32
},
{
"feature": "morgan_322",
"gain_proxy": 28
},
{
"feature": "morgan_314",
"gain_proxy": 27
},
{
"feature": "morgan_516",
"gain_proxy": 24
},
{
"feature": "morgan_833",
"gain_proxy": 24
},
{
"feature": "morgan_979",
"gain_proxy": 24
},
{
"feature": "morgan_351",
"gain_proxy": 21
},
{
"feature": "morgan_575",
"gain_proxy": 21
},
{
"feature": "morgan_582",
"gain_proxy": 21
}
],
"fit_and_eval_seconds": 0.5651998839966836
},
{
"name": "protein_aac_dipeptide_lightgbm",
"best_iteration": 65,
"validation": {
"rmse": 1.5175224729443033,
"mae": 1.1705128204333943,
"pearson_r": 0.43913909656574024,
"spearman_r": 0.4573197112813825,
"r2": 0.1645636422606147
},
"test": {
"rmse": 1.7425222015081137,
"mae": 1.4758731703045813,
"pearson_r": 0.28018428339607077,
"spearman_r": 0.20346550056123705,
"r2": 0.012335077998096189
},
"top_feature_importance": [
{
"feature": "dipeptide_DV",
"gain_proxy": 35
},
{
"feature": "dipeptide_LM",
"gain_proxy": 34
},
{
"feature": "dipeptide_CW",
"gain_proxy": 31
},
{
"feature": "dipeptide_II",
"gain_proxy": 29
},
{
"feature": "dipeptide_SQ",
"gain_proxy": 29
},
{
"feature": "dipeptide_RN",
"gain_proxy": 27
},
{
"feature": "dipeptide_SK",
"gain_proxy": 27
},
{
"feature": "dipeptide_NL",
"gain_proxy": 24
},
{
"feature": "dipeptide_PS",
"gain_proxy": 24
},
{
"feature": "dipeptide_DP",
"gain_proxy": 22
},
{
"feature": "dipeptide_AE",
"gain_proxy": 21
},
{
"feature": "dipeptide_AK",
"gain_proxy": 21
},
{
"feature": "dipeptide_FG",
"gain_proxy": 21
},
{
"feature": "dipeptide_GS",
"gain_proxy": 21
},
{
"feature": "log_protein_length",
"gain_proxy": 20
},
{
"feature": "dipeptide_ME",
"gain_proxy": 20
},
{
"feature": "dipeptide_YA",
"gain_proxy": 20
},
{
"feature": "aac_C",
"gain_proxy": 19
},
{
"feature": "aac_D",
"gain_proxy": 19
},
{
"feature": "aac_V",
"gain_proxy": 19
}
],
"fit_and_eval_seconds": 1.1578160819990444
},
{
"name": "combined_morgan_protein_lightgbm",
"best_iteration": 3,
"validation": {
"rmse": 1.6403288426299631,
"mae": 1.2758415765589668,
"pearson_r": 0.274395891588625,
"spearman_r": 0.11913397164880844,
"r2": 0.023876087895473352
},
"test": {
"rmse": 1.7417590913562446,
"mae": 1.461483008436187,
"pearson_r": 0.35160992407209807,
"spearman_r": 0.34828677658449575,
"r2": 0.013199953169361733
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 7
},
{
"feature": "dipeptide_II",
"gain_proxy": 3
},
{
"feature": "morgan_679",
"gain_proxy": 3
},
{
"feature": "dipeptide_AE",
"gain_proxy": 2
},
{
"feature": "dipeptide_EY",
"gain_proxy": 2
},
{
"feature": "dipeptide_KR",
"gain_proxy": 2
},
{
"feature": "dipeptide_LA",
"gain_proxy": 2
},
{
"feature": "dipeptide_SK",
"gain_proxy": 2
},
{
"feature": "dipeptide_SL",
"gain_proxy": 2
},
{
"feature": "dipeptide_VC",
"gain_proxy": 2
},
{
"feature": "dipeptide_VS",
"gain_proxy": 2
},
{
"feature": "morgan_1766",
"gain_proxy": 2
},
{
"feature": "morgan_1985",
"gain_proxy": 2
},
{
"feature": "aac_A",
"gain_proxy": 1
},
{
"feature": "aac_G",
"gain_proxy": 1
},
{
"feature": "aac_P",
"gain_proxy": 1
},
{
"feature": "aac_S",
"gain_proxy": 1
},
{
"feature": "aac_V",
"gain_proxy": 1
},
{
"feature": "dipeptide_AL",
"gain_proxy": 1
},
{
"feature": "dipeptide_AQ",
"gain_proxy": 1
}
],
"fit_and_eval_seconds": 0.6484794639982283
}
]
},
"cold_scaffold": {
"rows": {
"train": 1863,
"validation": 222,
"test": 242
},
"mean_baseline": {
"prediction": 6.359398260690221,
"validation": {
"rmse": 1.3619980768982987,
"mae": 1.1612610862511097,
"pearson_r": null,
"spearman_r": null,
"r2": -0.006402567332228326
},
"test": {
"rmse": 1.6697649191739543,
"mae": 1.4104975620384284,
"pearson_r": null,
"spearman_r": null,
"r2": -0.084905858576247
}
},
"models": [
{
"name": "ligand_morgan_descriptors_lightgbm",
"best_iteration": 13,
"validation": {
"rmse": 1.2958726238446427,
"mae": 1.154764312318227,
"pearson_r": 0.37287543649042026,
"spearman_r": 0.29801265103744795,
"r2": 0.08894755565243462
},
"test": {
"rmse": 1.523944051676444,
"mae": 1.2685533732962604,
"pearson_r": 0.5183185822284528,
"spearman_r": 0.5410297865375379,
"r2": 0.09631008516280415
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 29
},
{
"feature": "ring_count",
"gain_proxy": 21
},
{
"feature": "mol_logp",
"gain_proxy": 20
},
{
"feature": "tpsa",
"gain_proxy": 18
},
{
"feature": "fraction_csp3",
"gain_proxy": 17
},
{
"feature": "morgan_351",
"gain_proxy": 12
},
{
"feature": "morgan_1452",
"gain_proxy": 11
},
{
"feature": "morgan_1607",
"gain_proxy": 10
},
{
"feature": "h_donors",
"gain_proxy": 9
},
{
"feature": "morgan_833",
"gain_proxy": 8
},
{
"feature": "morgan_1126",
"gain_proxy": 7
},
{
"feature": "morgan_1961",
"gain_proxy": 7
},
{
"feature": "rotatable_bonds",
"gain_proxy": 7
},
{
"feature": "morgan_392",
"gain_proxy": 6
},
{
"feature": "morgan_1104",
"gain_proxy": 6
},
{
"feature": "morgan_516",
"gain_proxy": 5
},
{
"feature": "morgan_953",
"gain_proxy": 5
},
{
"feature": "h_acceptors",
"gain_proxy": 5
},
{
"feature": "morgan_118",
"gain_proxy": 4
},
{
"feature": "morgan_283",
"gain_proxy": 4
}
],
"fit_and_eval_seconds": 0.2359873610039358
},
{
"name": "protein_aac_dipeptide_lightgbm",
"best_iteration": 31,
"validation": {
"rmse": 1.2306961672737913,
"mae": 1.0710334695147479,
"pearson_r": 0.4402069153842137,
"spearman_r": 0.4076833999544298,
"r2": 0.17828645962908496
},
"test": {
"rmse": 1.3907344532488832,
"mae": 1.1226538845330731,
"pearson_r": 0.549348958831238,
"spearman_r": 0.5113706394958901,
"r2": 0.2473902905666937
},
"top_feature_importance": [
{
"feature": "dipeptide_KG",
"gain_proxy": 21
},
{
"feature": "dipeptide_II",
"gain_proxy": 20
},
{
"feature": "dipeptide_MF",
"gain_proxy": 14
},
{
"feature": "dipeptide_PE",
"gain_proxy": 14
},
{
"feature": "dipeptide_SK",
"gain_proxy": 14
},
{
"feature": "dipeptide_KD",
"gain_proxy": 13
},
{
"feature": "dipeptide_KY",
"gain_proxy": 13
},
{
"feature": "dipeptide_ME",
"gain_proxy": 13
},
{
"feature": "aac_G",
"gain_proxy": 12
},
{
"feature": "dipeptide_PG",
"gain_proxy": 12
},
{
"feature": "dipeptide_PR",
"gain_proxy": 12
},
{
"feature": "dipeptide_QE",
"gain_proxy": 12
},
{
"feature": "aac_V",
"gain_proxy": 11
},
{
"feature": "dipeptide_CW",
"gain_proxy": 11
},
{
"feature": "dipeptide_TE",
"gain_proxy": 11
},
{
"feature": "dipeptide_VS",
"gain_proxy": 11
},
{
"feature": "dipeptide_GL",
"gain_proxy": 10
},
{
"feature": "dipeptide_PV",
"gain_proxy": 10
},
{
"feature": "dipeptide_KE",
"gain_proxy": 9
},
{
"feature": "dipeptide_LT",
"gain_proxy": 9
}
],
"fit_and_eval_seconds": 0.5335308360008639
},
{
"name": "combined_morgan_protein_lightgbm",
"best_iteration": 34,
"validation": {
"rmse": 1.2622611214138577,
"mae": 1.1051660382647568,
"pearson_r": 0.401625369186114,
"spearman_r": 0.2659324149990984,
"r2": 0.13559522174943062
},
"test": {
"rmse": 1.3120002515276956,
"mae": 1.0707845139526735,
"pearson_r": 0.6537000142994885,
"spearman_r": 0.6386787033375942,
"r2": 0.33019370323802355
},
"top_feature_importance": [
{
"feature": "ring_count",
"gain_proxy": 36
},
{
"feature": "dipeptide_II",
"gain_proxy": 34
},
{
"feature": "mol_weight",
"gain_proxy": 31
},
{
"feature": "aac_G",
"gain_proxy": 21
},
{
"feature": "dipeptide_DG",
"gain_proxy": 20
},
{
"feature": "rotatable_bonds",
"gain_proxy": 20
},
{
"feature": "tpsa",
"gain_proxy": 18
},
{
"feature": "dipeptide_HW",
"gain_proxy": 14
},
{
"feature": "dipeptide_PR",
"gain_proxy": 14
},
{
"feature": "dipeptide_ME",
"gain_proxy": 13
},
{
"feature": "aac_Y",
"gain_proxy": 12
},
{
"feature": "dipeptide_CW",
"gain_proxy": 12
},
{
"feature": "dipeptide_GL",
"gain_proxy": 12
},
{
"feature": "dipeptide_GP",
"gain_proxy": 11
},
{
"feature": "dipeptide_TN",
"gain_proxy": 11
},
{
"feature": "mol_logp",
"gain_proxy": 11
},
{
"feature": "dipeptide_LF",
"gain_proxy": 10
},
{
"feature": "fraction_csp3",
"gain_proxy": 10
},
{
"feature": "dipeptide_AD",
"gain_proxy": 9
},
{
"feature": "dipeptide_AV",
"gain_proxy": 9
}
],
"fit_and_eval_seconds": 1.0606267510011094
}
]
},
"protein_similarity_50": {
"rows": {
"train": 1879,
"validation": 222,
"test": 226
},
"mean_baseline": {
"prediction": 6.197894919416744,
"validation": {
"rmse": 1.587711147693772,
"mae": 1.3285350835390142,
"pearson_r": null,
"spearman_r": null,
"r2": -0.3168171166313214
},
"test": {
"rmse": 1.9936052456701605,
"mae": 1.6715350127460604,
"pearson_r": null,
"spearman_r": null,
"r2": -0.06840343733542187
}
},
"models": [
{
"name": "ligand_morgan_descriptors_lightgbm",
"best_iteration": 1,
"validation": {
"rmse": 1.5969844670996178,
"mae": 1.3332309895605403,
"pearson_r": 0.2203253411571658,
"spearman_r": 0.16356024012516915,
"r2": -0.33224426383343597
},
"test": {
"rmse": 1.977604347343066,
"mae": 1.658904212276055,
"pearson_r": 0.4026196495659993,
"spearman_r": 0.40535654733382154,
"r2": -0.05132201165589212
},
"top_feature_importance": [
{
"feature": "fraction_csp3",
"gain_proxy": 3
},
{
"feature": "mol_weight",
"gain_proxy": 2
},
{
"feature": "mol_logp",
"gain_proxy": 2
},
{
"feature": "tpsa",
"gain_proxy": 2
},
{
"feature": "morgan_89",
"gain_proxy": 1
},
{
"feature": "morgan_242",
"gain_proxy": 1
},
{
"feature": "morgan_351",
"gain_proxy": 1
},
{
"feature": "morgan_383",
"gain_proxy": 1
},
{
"feature": "morgan_694",
"gain_proxy": 1
},
{
"feature": "morgan_1104",
"gain_proxy": 1
},
{
"feature": "morgan_1129",
"gain_proxy": 1
},
{
"feature": "morgan_1136",
"gain_proxy": 1
},
{
"feature": "morgan_1164",
"gain_proxy": 1
},
{
"feature": "morgan_1309",
"gain_proxy": 1
},
{
"feature": "morgan_1349",
"gain_proxy": 1
},
{
"feature": "morgan_1452",
"gain_proxy": 1
},
{
"feature": "morgan_1455",
"gain_proxy": 1
},
{
"feature": "morgan_1516",
"gain_proxy": 1
},
{
"feature": "morgan_1602",
"gain_proxy": 1
},
{
"feature": "morgan_1603",
"gain_proxy": 1
}
],
"fit_and_eval_seconds": 0.2233759930022643
},
{
"name": "protein_aac_dipeptide_lightgbm",
"best_iteration": 2,
"validation": {
"rmse": 1.5833372852629355,
"mae": 1.3266514496830648,
"pearson_r": 0.23599130325754594,
"spearman_r": 0.15471350052315208,
"r2": -0.3095719151134446
},
"test": {
"rmse": 1.9883842273036123,
"mae": 1.666134010195384,
"pearson_r": 0.10073591945417193,
"spearman_r": 0.03801325161297808,
"r2": -0.0628147183894443
},
"top_feature_importance": [
{
"feature": "dipeptide_RN",
"gain_proxy": 3
},
{
"feature": "aac_C",
"gain_proxy": 2
},
{
"feature": "aac_L",
"gain_proxy": 2
},
{
"feature": "dipeptide_AK",
"gain_proxy": 2
},
{
"feature": "dipeptide_EQ",
"gain_proxy": 2
},
{
"feature": "dipeptide_ER",
"gain_proxy": 2
},
{
"feature": "dipeptide_GV",
"gain_proxy": 2
},
{
"feature": "dipeptide_HA",
"gain_proxy": 2
},
{
"feature": "dipeptide_IG",
"gain_proxy": 2
},
{
"feature": "dipeptide_KM",
"gain_proxy": 2
},
{
"feature": "dipeptide_LL",
"gain_proxy": 2
},
{
"feature": "dipeptide_QL",
"gain_proxy": 2
},
{
"feature": "dipeptide_RP",
"gain_proxy": 2
},
{
"feature": "dipeptide_TK",
"gain_proxy": 2
},
{
"feature": "dipeptide_VV",
"gain_proxy": 2
},
{
"feature": "aac_D",
"gain_proxy": 1
},
{
"feature": "aac_E",
"gain_proxy": 1
},
{
"feature": "aac_Y",
"gain_proxy": 1
},
{
"feature": "dipeptide_AC",
"gain_proxy": 1
},
{
"feature": "dipeptide_AQ",
"gain_proxy": 1
}
],
"fit_and_eval_seconds": 0.30209682999702636
},
{
"name": "combined_morgan_protein_lightgbm",
"best_iteration": 1,
"validation": {
"rmse": 1.611691283951277,
"mae": 1.3438524054919534,
"pearson_r": -0.10804879245090941,
"spearman_r": -0.11406874130507588,
"r2": -0.35689483526187527
},
"test": {
"rmse": 1.9821888129918923,
"mae": 1.6608920028298857,
"pearson_r": 0.36332328596285074,
"spearman_r": 0.3483094176069358,
"r2": -0.0562019930396116
},
"top_feature_importance": [
{
"feature": "aac_V",
"gain_proxy": 2
},
{
"feature": "dipeptide_II",
"gain_proxy": 2
},
{
"feature": "fraction_csp3",
"gain_proxy": 2
},
{
"feature": "aac_I",
"gain_proxy": 1
},
{
"feature": "aac_L",
"gain_proxy": 1
},
{
"feature": "aac_R",
"gain_proxy": 1
},
{
"feature": "dipeptide_AI",
"gain_proxy": 1
},
{
"feature": "dipeptide_DD",
"gain_proxy": 1
},
{
"feature": "dipeptide_DG",
"gain_proxy": 1
},
{
"feature": "dipeptide_FV",
"gain_proxy": 1
},
{
"feature": "dipeptide_GD",
"gain_proxy": 1
},
{
"feature": "dipeptide_HL",
"gain_proxy": 1
},
{
"feature": "dipeptide_IA",
"gain_proxy": 1
},
{
"feature": "dipeptide_KV",
"gain_proxy": 1
},
{
"feature": "dipeptide_ME",
"gain_proxy": 1
},
{
"feature": "dipeptide_PE",
"gain_proxy": 1
},
{
"feature": "dipeptide_PR",
"gain_proxy": 1
},
{
"feature": "dipeptide_QH",
"gain_proxy": 1
},
{
"feature": "dipeptide_ST",
"gain_proxy": 1
},
{
"feature": "dipeptide_TE",
"gain_proxy": 1
}
],
"fit_and_eval_seconds": 0.6783155190059915
}
]
},
"publication_time": {
"rows": {
"train": 1975,
"validation": 168,
"test": 184
},
"mean_baseline": {
"prediction": 6.398630262507186,
"validation": {
"rmse": 1.646194028357668,
"mae": 1.419704700798333,
"pearson_r": null,
"spearman_r": null,
"r2": -0.1609680053374234
},
"test": {
"rmse": 1.59244950858557,
"mae": 1.3391793961350122,
"pearson_r": null,
"spearman_r": null,
"r2": -0.07467789280549031
}
},
"models": [
{
"name": "ligand_morgan_descriptors_lightgbm",
"best_iteration": 92,
"validation": {
"rmse": 1.4550611893791385,
"mae": 1.2018820878829315,
"pearson_r": 0.375443514621232,
"spearman_r": 0.3150410678662765,
"r2": 0.09297194251144758
},
"test": {
"rmse": 1.5133235515419585,
"mae": 1.2467986818612975,
"pearson_r": 0.7114083803946392,
"spearman_r": 0.7234514563792849,
"r2": 0.02946644604098847
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 150
},
{
"feature": "tpsa",
"gain_proxy": 123
},
{
"feature": "mol_logp",
"gain_proxy": 106
},
{
"feature": "fraction_csp3",
"gain_proxy": 97
},
{
"feature": "ring_count",
"gain_proxy": 71
},
{
"feature": "rotatable_bonds",
"gain_proxy": 67
},
{
"feature": "h_acceptors",
"gain_proxy": 52
},
{
"feature": "morgan_1607",
"gain_proxy": 50
},
{
"feature": "h_donors",
"gain_proxy": 50
},
{
"feature": "morgan_1126",
"gain_proxy": 34
},
{
"feature": "morgan_1745",
"gain_proxy": 33
},
{
"feature": "morgan_114",
"gain_proxy": 32
},
{
"feature": "morgan_516",
"gain_proxy": 32
},
{
"feature": "morgan_582",
"gain_proxy": 27
},
{
"feature": "morgan_1114",
"gain_proxy": 26
},
{
"feature": "morgan_1956",
"gain_proxy": 26
},
{
"feature": "morgan_322",
"gain_proxy": 24
},
{
"feature": "morgan_1961",
"gain_proxy": 23
},
{
"feature": "morgan_446",
"gain_proxy": 22
},
{
"feature": "morgan_979",
"gain_proxy": 22
}
],
"fit_and_eval_seconds": 0.5143499049954698
},
{
"name": "protein_aac_dipeptide_lightgbm",
"best_iteration": 17,
"validation": {
"rmse": 1.5351615319490832,
"mae": 1.2969109325410964,
"pearson_r": 0.23528694230179825,
"spearman_r": 0.26719521624061504,
"r2": -0.009639576824195473
},
"test": {
"rmse": 1.5731180297912137,
"mae": 1.3334084079245385,
"pearson_r": -0.44804424444095664,
"spearman_r": -0.4685459080574742,
"r2": -0.0487442434844505
},
"top_feature_importance": [
{
"feature": "dipeptide_FL",
"gain_proxy": 14
},
{
"feature": "dipeptide_KV",
"gain_proxy": 14
},
{
"feature": "dipeptide_GV",
"gain_proxy": 11
},
{
"feature": "dipeptide_IL",
"gain_proxy": 11
},
{
"feature": "dipeptide_MP",
"gain_proxy": 11
},
{
"feature": "dipeptide_DA",
"gain_proxy": 9
},
{
"feature": "dipeptide_PA",
"gain_proxy": 9
},
{
"feature": "aac_G",
"gain_proxy": 8
},
{
"feature": "dipeptide_GD",
"gain_proxy": 8
},
{
"feature": "dipeptide_KR",
"gain_proxy": 8
},
{
"feature": "dipeptide_AE",
"gain_proxy": 7
},
{
"feature": "dipeptide_DE",
"gain_proxy": 7
},
{
"feature": "dipeptide_VA",
"gain_proxy": 7
},
{
"feature": "dipeptide_IE",
"gain_proxy": 6
},
{
"feature": "dipeptide_PS",
"gain_proxy": 6
},
{
"feature": "dipeptide_VW",
"gain_proxy": 6
},
{
"feature": "aac_Q",
"gain_proxy": 5
},
{
"feature": "aac_W",
"gain_proxy": 5
},
{
"feature": "dipeptide_CW",
"gain_proxy": 5
},
{
"feature": "dipeptide_EK",
"gain_proxy": 5
}
],
"fit_and_eval_seconds": 0.43806035300076474
},
{
"name": "combined_morgan_protein_lightgbm",
"best_iteration": 200,
"validation": {
"rmse": 1.380413857452562,
"mae": 1.1233174497640375,
"pearson_r": 0.4438161613542017,
"spearman_r": 0.4131731753053455,
"r2": 0.1836491896186765
},
"test": {
"rmse": 1.6254562021089098,
"mae": 1.3579855855345364,
"pearson_r": 0.42985292795295293,
"spearman_r": 0.3108117025213188,
"r2": -0.11968927071096136
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 252
},
{
"feature": "mol_logp",
"gain_proxy": 151
},
{
"feature": "tpsa",
"gain_proxy": 115
},
{
"feature": "fraction_csp3",
"gain_proxy": 90
},
{
"feature": "rotatable_bonds",
"gain_proxy": 66
},
{
"feature": "ring_count",
"gain_proxy": 55
},
{
"feature": "morgan_347",
"gain_proxy": 54
},
{
"feature": "dipeptide_GL",
"gain_proxy": 50
},
{
"feature": "dipeptide_RN",
"gain_proxy": 42
},
{
"feature": "h_acceptors",
"gain_proxy": 42
},
{
"feature": "morgan_679",
"gain_proxy": 38
},
{
"feature": "aac_C",
"gain_proxy": 37
},
{
"feature": "h_donors",
"gain_proxy": 37
},
{
"feature": "morgan_108",
"gain_proxy": 34
},
{
"feature": "aac_W",
"gain_proxy": 33
},
{
"feature": "morgan_1766",
"gain_proxy": 31
},
{
"feature": "aac_T",
"gain_proxy": 30
},
{
"feature": "dipeptide_CW",
"gain_proxy": 30
},
{
"feature": "dipeptide_ME",
"gain_proxy": 30
},
{
"feature": "morgan_125",
"gain_proxy": 30
}
],
"fit_and_eval_seconds": 2.884516165002424
}
]
},
"random_pair": {
"rows": {
"train": 1864,
"validation": 233,
"test": 230
},
"mean_baseline": {
"prediction": 6.373521205197287,
"validation": {
"rmse": 1.6661731479367692,
"mae": 1.3803133580560933,
"pearson_r": null,
"spearman_r": null,
"r2": -0.025274664404281832
},
"test": {
"rmse": 1.7689644621168779,
"mae": 1.5146153453218127,
"pearson_r": null,
"spearman_r": null,
"r2": -0.022869546423538667
}
},
"models": [
{
"name": "ligand_morgan_descriptors_lightgbm",
"best_iteration": 291,
"validation": {
"rmse": 0.890881110910472,
"mae": 0.6780673290669472,
"pearson_r": 0.8450300803577628,
"spearman_r": 0.8382415234734463,
"r2": 0.706884040738585
},
"test": {
"rmse": 0.9089204637589584,
"mae": 0.6910445846855048,
"pearson_r": 0.8556939615450855,
"spearman_r": 0.835809968779844,
"r2": 0.7299564583761792
},
"top_feature_importance": [
{
"feature": "mol_weight",
"gain_proxy": 544
},
{
"feature": "mol_logp",
"gain_proxy": 515
},
{
"feature": "tpsa",
"gain_proxy": 434
},
{
"feature": "fraction_csp3",
"gain_proxy": 431
},
{
"feature": "rotatable_bonds",
"gain_proxy": 186
},
{
"feature": "h_acceptors",
"gain_proxy": 150
},
{
"feature": "ring_count",
"gain_proxy": 126
},
{
"feature": "h_donors",
"gain_proxy": 112
},
{
"feature": "morgan_1607",
"gain_proxy": 70
},
{
"feature": "morgan_1114",
"gain_proxy": 65
},
{
"feature": "morgan_979",
"gain_proxy": 64
},
{
"feature": "morgan_891",
"gain_proxy": 54
},
{
"feature": "morgan_1750",
"gain_proxy": 52
},
{
"feature": "morgan_1295",
"gain_proxy": 51
},
{
"feature": "morgan_1679",
"gain_proxy": 51
},
{
"feature": "morgan_1745",
"gain_proxy": 50
},
{
"feature": "morgan_354",
"gain_proxy": 49
},
{
"feature": "morgan_1860",
"gain_proxy": 48
},
{
"feature": "morgan_516",
"gain_proxy": 47
},
{
"feature": "morgan_935",
"gain_proxy": 46
}
],
"fit_and_eval_seconds": 1.0006999510005699
},
{
"name": "protein_aac_dipeptide_lightgbm",
"best_iteration": 79,
"validation": {
"rmse": 1.1239582999352296,
"mae": 0.8451331462113194,
"pearson_r": 0.7313415132801145,
"spearman_r": 0.7218691728662899,
"r2": 0.5334476718747884
},
"test": {
"rmse": 1.0689895424167133,
"mae": 0.82602575692832,
"pearson_r": 0.798450172465141,
"spearman_r": 0.7827748318910662,
"r2": 0.6264670259596485
},
"top_feature_importance": [
{
"feature": "dipeptide_II",
"gain_proxy": 55
},
{
"feature": "dipeptide_CW",
"gain_proxy": 42
},
{
"feature": "aac_G",
"gain_proxy": 30
},
{
"feature": "dipeptide_SK",
"gain_proxy": 29
},
{
"feature": "aac_V",
"gain_proxy": 26
},
{
"feature": "dipeptide_IN",
"gain_proxy": 26
},
{
"feature": "dipeptide_MF",
"gain_proxy": 26
},
{
"feature": "dipeptide_PE",
"gain_proxy": 26
},
{
"feature": "dipeptide_PG",
"gain_proxy": 26
},
{
"feature": "dipeptide_TE",
"gain_proxy": 25
},
{
"feature": "dipeptide_KM",
"gain_proxy": 24
},
{
"feature": "aac_C",
"gain_proxy": 23
},
{
"feature": "dipeptide_KE",
"gain_proxy": 22
},
{
"feature": "aac_H",
"gain_proxy": 21
},
{
"feature": "dipeptide_GL",
"gain_proxy": 21
},
{
"feature": "dipeptide_QL",
"gain_proxy": 19
},
{
"feature": "dipeptide_VC",
"gain_proxy": 19
},
{
"feature": "dipeptide_AF",
"gain_proxy": 18
},
{
"feature": "dipeptide_GV",
"gain_proxy": 18
},
{
"feature": "aac_A",
"gain_proxy": 17
}
],
"fit_and_eval_seconds": 0.8802312090047053
},
{
"name": "combined_morgan_protein_lightgbm",
"best_iteration": 589,
"validation": {
"rmse": 0.7847182262329421,
"mae": 0.5505910287586331,
"pearson_r": 0.8818635635033661,
"spearman_r": 0.8802811373240753,
"r2": 0.7725806346888339
},
"test": {
"rmse": 0.6816113347658449,
"mae": 0.47507585180354917,
"pearson_r": 0.921295522373731,
"spearman_r": 0.8971091578763459,
"r2": 0.8481357183897692
},
"top_feature_importance": [
{
"feature": "mol_logp",
"gain_proxy": 839
},
{
"feature": "mol_weight",
"gain_proxy": 778
},
{
"feature": "tpsa",
"gain_proxy": 430
},
{
"feature": "fraction_csp3",
"gain_proxy": 417
},
{
"feature": "rotatable_bonds",
"gain_proxy": 253
},
{
"feature": "h_acceptors",
"gain_proxy": 171
},
{
"feature": "dipeptide_TE",
"gain_proxy": 139
},
{
"feature": "morgan_108",
"gain_proxy": 120
},
{
"feature": "dipeptide_IT",
"gain_proxy": 115
},
{
"feature": "dipeptide_II",
"gain_proxy": 113
},
{
"feature": "ring_count",
"gain_proxy": 112
},
{
"feature": "h_donors",
"gain_proxy": 90
},
{
"feature": "morgan_1326",
"gain_proxy": 87
},
{
"feature": "aac_T",
"gain_proxy": 84
},
{
"feature": "morgan_533",
"gain_proxy": 79
},
{
"feature": "dipeptide_GL",
"gain_proxy": 77
},
{
"feature": "aac_A",
"gain_proxy": 74
},
{
"feature": "morgan_125",
"gain_proxy": 73
},
{
"feature": "morgan_1411",
"gain_proxy": 71
},
{
"feature": "dipeptide_NV",
"gain_proxy": 67
}
],
"fit_and_eval_seconds": 6.179297015005432
}
]
}
}
}