Concise copy + em-dash removal + calibration on leaderboard

website/src/App.tsx

@@ -55,8 +55,7 @@ export default function App() {
       <footer className="footer">
         <div className="footer-inner">
           <span>
-            HydroPD — MS detectability prediction for non-animal food-protein
-            hydrolysates
+            HydroPD. MS detectability for non-animal protein hydrolysates.
           </span>
           <span>
             <a

website/src/components/ConditionMatcher.tsx

@@ -40,8 +40,7 @@ export default function ConditionMatcher({
       <div className="field-label">
         Match my conditions{' '}
         <span className="field-hint">
-          Describe how your peptides were made — we rank the models by similarity
-          (species → enzyme → protocol).
+          How were your peptides made? We rank models by species, then enzyme, then protocol.
         </span>
       </div>
 

website/src/data/generated/model_leaderboard.json

@@ -7,7 +7,9 @@
   "f1",
   "precision",
   "recall",
-  "specificity"
+  "specificity",
+  "brier",
+  "ece"
  ],
  "metricLabels": {
   "auroc": "AUROC",
@@ -17,11 +19,13 @@
   "f1": "F1",
   "precision": "Precision",
   "recall": "Sensitivity / recall",
-  "specificity": "Specificity"
+  "specificity": "Specificity",
+  "brier": "Brier",
+  "ece": "ECE"
  },
  "nSpecies": 8,
  "nDatasets": 30,
- "note": "Aggregated across species from 5-fold protein-grouped CV on the 30 usable datasets. Per-metric values are the mean over folds, then over a species' datasets, then across species; min/max identify the weakest/strongest species.",
+ "note": "Aggregated across species from 5-fold protein-grouped CV on the 30 usable datasets. Each value is the mean over folds, then over a species' datasets, then across species. Brier and ECE are calibration metrics where lower is better.",
  "models": [
   {
    "id": "pfly_ft",
@@ -106,7 +110,7 @@
    "classifier": "xgb",
    "featureSet": "pep_phys",
    "label": "XGBoost \u00b7 PepBERT + physicochemical",
-   "description": "Gradient-boosted decision trees (XGBoost) \u2014 the most consistent winner across datasets. PepBERT embeddings concatenated with the 14 physicochemical descriptors.",
+   "description": "Gradient-boosted decision trees. The most consistent winner across datasets. PepBERT embeddings concatenated with the 14 physicochemical descriptors.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -173,6 +177,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.186,
+     "min": 0.1178,
+     "max": 0.2698,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1555,
+     "min": 0.0626,
+     "max": 0.2674,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -186,7 +206,9 @@
      "f1": 0.8409,
      "precision": 0.818,
      "recall": 0.8682,
-     "specificity": 0.7887
+     "specificity": 0.7887,
+     "brier": 0.1178,
+     "ece": 0.0951
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -198,7 +220,9 @@
      "f1": 0.8249,
      "precision": 0.8005,
      "recall": 0.8569,
-     "specificity": 0.7778
+     "specificity": 0.7778,
+     "brier": 0.1394,
+     "ece": 0.1071
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -210,7 +234,9 @@
      "f1": 0.7936,
      "precision": 0.7814,
      "recall": 0.8093,
-     "specificity": 0.779
+     "specificity": 0.779,
+     "brier": 0.1469,
+     "ece": 0.0626
     },
     {
      "species": "Soybean (Glycine max)",
@@ -222,7 +248,9 @@
      "f1": 0.7349,
      "precision": 0.7065,
      "recall": 0.8187,
-     "specificity": 0.7451
+     "specificity": 0.7451,
+     "brier": 0.1834,
+     "ece": 0.1672
     },
     {
      "species": "wheat",
@@ -234,7 +262,9 @@
      "f1": 0.7695,
      "precision": 0.7272,
      "recall": 0.8239,
-     "specificity": 0.6711
+     "specificity": 0.6711,
+     "brier": 0.1809,
+     "ece": 0.1174
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -246,7 +276,9 @@
      "f1": 0.7278,
      "precision": 0.6995,
      "recall": 0.7826,
-     "specificity": 0.5952
+     "specificity": 0.5952,
+     "brier": 0.2226,
+     "ece": 0.2067
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -258,7 +290,9 @@
      "f1": 0.7062,
      "precision": 0.716,
      "recall": 0.7314,
-     "specificity": 0.6509
+     "specificity": 0.6509,
+     "brier": 0.2272,
+     "ece": 0.2209
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -270,7 +304,9 @@
      "f1": 0.7302,
      "precision": 0.7042,
      "recall": 0.8037,
-     "specificity": 0.4583
+     "specificity": 0.4583,
+     "brier": 0.2698,
+     "ece": 0.2674
     }
    ],
    "partial": false
@@ -280,7 +316,7 @@
    "classifier": "xgb",
    "featureSet": "esm2_phys",
    "label": "XGBoost \u00b7 ESM2 + physicochemical",
-   "description": "Gradient-boosted decision trees (XGBoost) \u2014 the most consistent winner across datasets. ESM2 embeddings concatenated with the 14 physicochemical descriptors \u2014 best overall.",
+   "description": "Gradient-boosted decision trees. The most consistent winner across datasets. ESM2 embeddings plus the 14 physicochemical descriptors. Best overall.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -347,6 +383,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1918,
+     "min": 0.1168,
+     "max": 0.28,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1594,
+     "min": 0.0684,
+     "max": 0.2359,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -360,7 +412,9 @@
      "f1": 0.8377,
      "precision": 0.8395,
      "recall": 0.8406,
-     "specificity": 0.8315
+     "specificity": 0.8315,
+     "brier": 0.1168,
+     "ece": 0.1025
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -372,7 +426,9 @@
      "f1": 0.8161,
      "precision": 0.8009,
      "recall": 0.8394,
-     "specificity": 0.7867
+     "specificity": 0.7867,
+     "brier": 0.1454,
+     "ece": 0.1218
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -384,7 +440,9 @@
      "f1": 0.7895,
      "precision": 0.7781,
      "recall": 0.804,
-     "specificity": 0.7759
+     "specificity": 0.7759,
+     "brier": 0.1492,
+     "ece": 0.0684
     },
     {
      "species": "Soybean (Glycine max)",
@@ -396,7 +454,9 @@
      "f1": 0.7253,
      "precision": 0.7045,
      "recall": 0.7994,
-     "specificity": 0.7512
+     "specificity": 0.7512,
+     "brier": 0.1795,
+     "ece": 0.1552
     },
     {
      "species": "wheat",
@@ -408,7 +468,9 @@
      "f1": 0.7691,
      "precision": 0.7208,
      "recall": 0.8322,
-     "specificity": 0.6561
+     "specificity": 0.6561,
+     "brier": 0.1854,
+     "ece": 0.1304
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -420,7 +482,9 @@
      "f1": 0.7492,
      "precision": 0.7059,
      "recall": 0.8477,
-     "specificity": 0.4192
+     "specificity": 0.4192,
+     "brier": 0.28,
+     "ece": 0.2359
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -432,7 +496,9 @@
      "f1": 0.6923,
      "precision": 0.703,
      "recall": 0.7307,
-     "specificity": 0.6084
+     "specificity": 0.6084,
+     "brier": 0.2279,
+     "ece": 0.2338
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -444,7 +510,9 @@
      "f1": 0.6993,
      "precision": 0.6735,
      "recall": 0.76,
-     "specificity": 0.5337
+     "specificity": 0.5337,
+     "brier": 0.2506,
+     "ece": 0.227
     }
    ],
    "partial": false
@@ -454,7 +522,7 @@
    "classifier": "xgb",
    "featureSet": "esm2",
    "label": "XGBoost \u00b7 ESM2",
-   "description": "Gradient-boosted decision trees (XGBoost) \u2014 the most consistent winner across datasets. ESM2-t12-35M protein-language-model embeddings of the peptide sequence.",
+   "description": "Gradient-boosted decision trees. The most consistent winner across datasets. ESM2-t12-35M protein-language-model embeddings of the peptide sequence.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -521,6 +589,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1954,
+     "min": 0.1208,
+     "max": 0.2773,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1604,
+     "min": 0.0685,
+     "max": 0.2511,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Red rice bran (Oryza sativa)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -534,7 +618,9 @@
      "f1": 0.8437,
      "precision": 0.8542,
      "recall": 0.8348,
-     "specificity": 0.8482
+     "specificity": 0.8482,
+     "brier": 0.1208,
+     "ece": 0.102
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -546,7 +632,9 @@
      "f1": 0.8062,
      "precision": 0.7921,
      "recall": 0.8305,
-     "specificity": 0.7758
+     "specificity": 0.7758,
+     "brier": 0.1477,
+     "ece": 0.1146
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -558,7 +646,9 @@
      "f1": 0.7876,
      "precision": 0.7791,
      "recall": 0.7993,
-     "specificity": 0.779
+     "specificity": 0.779,
+     "brier": 0.1504,
+     "ece": 0.0685
     },
     {
      "species": "Soybean (Glycine max)",
@@ -570,7 +660,9 @@
      "f1": 0.7316,
      "precision": 0.7042,
      "recall": 0.8162,
-     "specificity": 0.7473
+     "specificity": 0.7473,
+     "brier": 0.1791,
+     "ece": 0.1557
     },
     {
      "species": "wheat",
@@ -582,7 +674,9 @@
      "f1": 0.7676,
      "precision": 0.7208,
      "recall": 0.8298,
-     "specificity": 0.6573
+     "specificity": 0.6573,
+     "brier": 0.1862,
+     "ece": 0.1316
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -594,7 +688,9 @@
      "f1": 0.8437,
      "precision": 0.8045,
      "recall": 0.9401,
-     "specificity": 0.4138
+     "specificity": 0.4138,
+     "brier": 0.2773,
+     "ece": 0.2214
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -606,7 +702,9 @@
      "f1": 0.6943,
      "precision": 0.7133,
      "recall": 0.7203,
-     "specificity": 0.6343
+     "specificity": 0.6343,
+     "brier": 0.235,
+     "ece": 0.2382
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -618,7 +716,9 @@
      "f1": 0.682,
      "precision": 0.6612,
      "recall": 0.7473,
-     "specificity": 0.5105
+     "specificity": 0.5105,
+     "brier": 0.2665,
+     "ece": 0.2511
     }
    ],
    "partial": false
@@ -628,7 +728,7 @@
    "classifier": "mlp",
    "featureSet": "esm2_phys",
    "label": "Neural net (MLP) \u00b7 ESM2 + physicochemical",
-   "description": "A small multilayer-perceptron neural net over the feature vector. ESM2 embeddings concatenated with the 14 physicochemical descriptors \u2014 best overall.",
+   "description": "A small multilayer-perceptron neural net over the feature vector. ESM2 embeddings plus the 14 physicochemical descriptors. Best overall.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -695,6 +795,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "Spirulina (Arthrospira platensis)",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.2029,
+     "min": 0.1327,
+     "max": 0.2769,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Red rice bran (Oryza sativa)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1802,
+     "min": 0.1191,
+     "max": 0.2531,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Red rice bran (Oryza sativa)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -708,7 +824,9 @@
      "f1": 0.8312,
      "precision": 0.8172,
      "recall": 0.8486,
-     "specificity": 0.7985
+     "specificity": 0.7985,
+     "brier": 0.1327,
+     "ece": 0.1191
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -720,7 +838,9 @@
      "f1": 0.8057,
      "precision": 0.8098,
      "recall": 0.8099,
-     "specificity": 0.8039
+     "specificity": 0.8039,
+     "brier": 0.153,
+     "ece": 0.1241
     },
     {
      "species": "Soybean (Glycine max)",
@@ -732,7 +852,9 @@
      "f1": 0.7471,
      "precision": 0.7302,
      "recall": 0.8094,
-     "specificity": 0.7795
+     "specificity": 0.7795,
+     "brier": 0.1722,
+     "ece": 0.1561
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -744,7 +866,9 @@
      "f1": 0.7797,
      "precision": 0.7812,
      "recall": 0.781,
-     "specificity": 0.7872
+     "specificity": 0.7872,
+     "brier": 0.1804,
+     "ece": 0.1544
     },
     {
      "species": "wheat",
@@ -756,7 +880,9 @@
      "f1": 0.7729,
      "precision": 0.7312,
      "recall": 0.8273,
-     "specificity": 0.6748
+     "specificity": 0.6748,
+     "brier": 0.2017,
+     "ece": 0.175
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -768,7 +894,9 @@
      "f1": 0.7464,
      "precision": 0.7107,
      "recall": 0.8304,
-     "specificity": 0.462
+     "specificity": 0.462,
+     "brier": 0.2747,
+     "ece": 0.237
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -780,7 +908,9 @@
      "f1": 0.682,
      "precision": 0.6966,
      "recall": 0.721,
-     "specificity": 0.6002
+     "specificity": 0.6002,
+     "brier": 0.232,
+     "ece": 0.2226
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -792,7 +922,9 @@
      "f1": 0.6743,
      "precision": 0.6503,
      "recall": 0.7413,
-     "specificity": 0.4788
+     "specificity": 0.4788,
+     "brier": 0.2769,
+     "ece": 0.2531
     }
    ],
    "partial": false
@@ -869,6 +1001,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.2006,
+     "min": 0.1231,
+     "max": 0.2925,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1798,
+     "min": 0.1037,
+     "max": 0.2584,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -882,7 +1030,9 @@
      "f1": 0.8367,
      "precision": 0.8287,
      "recall": 0.8483,
-     "specificity": 0.8114
+     "specificity": 0.8114,
+     "brier": 0.1231,
+     "ece": 0.1037
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -894,7 +1044,9 @@
      "f1": 0.8065,
      "precision": 0.81,
      "recall": 0.8144,
-     "specificity": 0.8011
+     "specificity": 0.8011,
+     "brier": 0.1563,
+     "ece": 0.1306
     },
     {
      "species": "Soybean (Glycine max)",
@@ -906,7 +1058,9 @@
      "f1": 0.7636,
      "precision": 0.7428,
      "recall": 0.8383,
-     "specificity": 0.7812
+     "specificity": 0.7812,
+     "brier": 0.1703,
+     "ece": 0.1648
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -918,7 +1072,9 @@
      "f1": 0.7786,
      "precision": 0.7812,
      "recall": 0.7791,
-     "specificity": 0.7874
+     "specificity": 0.7874,
+     "brier": 0.1831,
+     "ece": 0.1593
     },
     {
      "species": "wheat",
@@ -930,7 +1086,9 @@
      "f1": 0.765,
      "precision": 0.7259,
      "recall": 0.8171,
-     "specificity": 0.6682
+     "specificity": 0.6682,
+     "brier": 0.2043,
+     "ece": 0.175
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -942,7 +1100,9 @@
      "f1": 0.7456,
      "precision": 0.7047,
      "recall": 0.8359,
-     "specificity": 0.4296
+     "specificity": 0.4296,
+     "brier": 0.2925,
+     "ece": 0.2584
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -954,7 +1114,9 @@
      "f1": 0.6799,
      "precision": 0.6978,
      "recall": 0.6933,
-     "specificity": 0.6218
+     "specificity": 0.6218,
+     "brier": 0.2167,
+     "ece": 0.2017
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -966,7 +1128,9 @@
      "f1": 0.6764,
      "precision": 0.6675,
      "recall": 0.7195,
-     "specificity": 0.5521
+     "specificity": 0.5521,
+     "brier": 0.2585,
+     "ece": 0.2447
     }
    ],
    "partial": false
@@ -1043,6 +1207,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.2076,
+     "min": 0.1189,
+     "max": 0.3179,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1898,
+     "min": 0.0905,
+     "max": 0.2942,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -1056,7 +1236,9 @@
      "f1": 0.8505,
      "precision": 0.8453,
      "recall": 0.8591,
-     "specificity": 0.8321
+     "specificity": 0.8321,
+     "brier": 0.1189,
+     "ece": 0.0905
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -1068,7 +1250,9 @@
      "f1": 0.8189,
      "precision": 0.8074,
      "recall": 0.8361,
-     "specificity": 0.7977
+     "specificity": 0.7977,
+     "brier": 0.1486,
+     "ece": 0.1304
     },
     {
      "species": "Soybean (Glycine max)",
@@ -1080,7 +1264,9 @@
      "f1": 0.733,
      "precision": 0.7256,
      "recall": 0.8035,
-     "specificity": 0.768
+     "specificity": 0.768,
+     "brier": 0.1893,
+     "ece": 0.1892
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -1092,7 +1278,9 @@
      "f1": 0.7783,
      "precision": 0.7763,
      "recall": 0.784,
-     "specificity": 0.7798
+     "specificity": 0.7798,
+     "brier": 0.1831,
+     "ece": 0.1578
     },
     {
      "species": "wheat",
@@ -1104,7 +1292,9 @@
      "f1": 0.7615,
      "precision": 0.7341,
      "recall": 0.7972,
-     "specificity": 0.6942
+     "specificity": 0.6942,
+     "brier": 0.1977,
+     "ece": 0.1594
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -1116,7 +1306,9 @@
      "f1": 0.6983,
      "precision": 0.6933,
      "recall": 0.7455,
-     "specificity": 0.584
+     "specificity": 0.584,
+     "brier": 0.2559,
+     "ece": 0.2624
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -1128,7 +1320,9 @@
      "f1": 0.6748,
      "precision": 0.7103,
      "recall": 0.6921,
-     "specificity": 0.6543
+     "specificity": 0.6543,
+     "brier": 0.2497,
+     "ece": 0.2349
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -1140,7 +1334,9 @@
      "f1": 0.72,
      "precision": 0.7012,
      "recall": 0.8006,
-     "specificity": 0.4236
+     "specificity": 0.4236,
+     "brier": 0.3179,
+     "ece": 0.2942
     }
    ],
    "partial": false
@@ -1150,7 +1346,7 @@
    "classifier": "rf",
    "featureSet": "pep_phys",
    "label": "Random forest \u00b7 PepBERT + physicochemical",
-   "description": "Random forest \u2014 bagged decision trees; robust but usually a step behind XGBoost. PepBERT embeddings concatenated with the 14 physicochemical descriptors.",
+   "description": "Bagged decision trees. Robust, usually a step behind XGBoost. PepBERT embeddings concatenated with the 14 physicochemical descriptors.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -1217,6 +1413,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1867,
+     "min": 0.1346,
+     "max": 0.2391,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.181,
+     "min": 0.111,
+     "max": 0.3165,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -1230,7 +1442,9 @@
      "f1": 0.8393,
      "precision": 0.8325,
      "recall": 0.8493,
-     "specificity": 0.8173
+     "specificity": 0.8173,
+     "brier": 0.1346,
+     "ece": 0.1433
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -1242,7 +1456,9 @@
      "f1": 0.8094,
      "precision": 0.7938,
      "recall": 0.8352,
-     "specificity": 0.7732
+     "specificity": 0.7732,
+     "brier": 0.1551,
+     "ece": 0.1395
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -1254,7 +1470,9 @@
      "f1": 0.7745,
      "precision": 0.7705,
      "recall": 0.7838,
-     "specificity": 0.7729
+     "specificity": 0.7729,
+     "brier": 0.1653,
+     "ece": 0.111
     },
     {
      "species": "Soybean (Glycine max)",
@@ -1266,7 +1484,9 @@
      "f1": 0.7182,
      "precision": 0.6867,
      "recall": 0.8085,
-     "specificity": 0.723
+     "specificity": 0.723,
+     "brier": 0.1781,
+     "ece": 0.185
     },
     {
      "species": "wheat",
@@ -1278,7 +1498,9 @@
      "f1": 0.7497,
      "precision": 0.7113,
      "recall": 0.8047,
-     "specificity": 0.651
+     "specificity": 0.651,
+     "brier": 0.1879,
+     "ece": 0.1169
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -1290,7 +1512,9 @@
      "f1": 0.6922,
      "precision": 0.6962,
      "recall": 0.7371,
-     "specificity": 0.5902
+     "specificity": 0.5902,
+     "brier": 0.211,
+     "ece": 0.204
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -1302,7 +1526,9 @@
      "f1": 0.6913,
      "precision": 0.7102,
      "recall": 0.7419,
-     "specificity": 0.6087
+     "specificity": 0.6087,
+     "brier": 0.2227,
+     "ece": 0.2316
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -1314,7 +1540,9 @@
      "f1": 0.7217,
      "precision": 0.6949,
      "recall": 0.8117,
-     "specificity": 0.3821
+     "specificity": 0.3821,
+     "brier": 0.2391,
+     "ece": 0.3165
     }
    ],
    "partial": false
@@ -1324,7 +1552,7 @@
    "classifier": "rf",
    "featureSet": "esm2_phys",
    "label": "Random forest \u00b7 ESM2 + physicochemical",
-   "description": "Random forest \u2014 bagged decision trees; robust but usually a step behind XGBoost. ESM2 embeddings concatenated with the 14 physicochemical descriptors \u2014 best overall.",
+   "description": "Bagged decision trees. Robust, usually a step behind XGBoost. ESM2 embeddings plus the 14 physicochemical descriptors. Best overall.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -1391,6 +1619,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1839,
+     "min": 0.1383,
+     "max": 0.232,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1642,
+     "min": 0.0897,
+     "max": 0.2851,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -1404,7 +1648,9 @@
      "f1": 0.8185,
      "precision": 0.8316,
      "recall": 0.8115,
-     "specificity": 0.829
+     "specificity": 0.829,
+     "brier": 0.1383,
+     "ece": 0.1139
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -1416,7 +1662,9 @@
      "f1": 0.8028,
      "precision": 0.7936,
      "recall": 0.8219,
-     "specificity": 0.7809
+     "specificity": 0.7809,
+     "brier": 0.1493,
+     "ece": 0.1254
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -1428,7 +1676,9 @@
      "f1": 0.7744,
      "precision": 0.7729,
      "recall": 0.7805,
-     "specificity": 0.7768
+     "specificity": 0.7768,
+     "brier": 0.1602,
+     "ece": 0.0897
     },
     {
      "species": "Soybean (Glycine max)",
@@ -1440,7 +1690,9 @@
      "f1": 0.7093,
      "precision": 0.7035,
      "recall": 0.7739,
-     "specificity": 0.7522
+     "specificity": 0.7522,
+     "brier": 0.1729,
+     "ece": 0.1762
     },
     {
      "species": "wheat",
@@ -1452,7 +1704,9 @@
      "f1": 0.7526,
      "precision": 0.7065,
      "recall": 0.8162,
-     "specificity": 0.6377
+     "specificity": 0.6377,
+     "brier": 0.1836,
+     "ece": 0.1071
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -1464,7 +1718,9 @@
      "f1": 0.7349,
      "precision": 0.7016,
      "recall": 0.8311,
-     "specificity": 0.391
+     "specificity": 0.391,
+     "brier": 0.232,
+     "ece": 0.2851
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -1476,7 +1732,9 @@
      "f1": 0.6834,
      "precision": 0.7126,
      "recall": 0.7116,
-     "specificity": 0.6356
+     "specificity": 0.6356,
+     "brier": 0.2139,
+     "ece": 0.2293
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -1488,7 +1746,9 @@
      "f1": 0.6812,
      "precision": 0.6718,
      "recall": 0.7442,
-     "specificity": 0.5299
+     "specificity": 0.5299,
+     "brier": 0.2206,
+     "ece": 0.187
     }
    ],
    "partial": false
@@ -1498,7 +1758,7 @@
    "classifier": "rf",
    "featureSet": "esm2",
    "label": "Random forest \u00b7 ESM2",
-   "description": "Random forest \u2014 bagged decision trees; robust but usually a step behind XGBoost. ESM2-t12-35M protein-language-model embeddings of the peptide sequence.",
+   "description": "Bagged decision trees. Robust, usually a step behind XGBoost. ESM2-t12-35M protein-language-model embeddings of the peptide sequence.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -1565,6 +1825,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1862,
+     "min": 0.1418,
+     "max": 0.2309,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1634,
+     "min": 0.0924,
+     "max": 0.2759,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -1578,7 +1854,9 @@
      "f1": 0.8064,
      "precision": 0.8091,
      "recall": 0.8092,
-     "specificity": 0.7869
+     "specificity": 0.7869,
+     "brier": 0.1418,
+     "ece": 0.1188
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -1590,7 +1868,9 @@
      "f1": 0.7992,
      "precision": 0.7952,
      "recall": 0.8139,
-     "specificity": 0.7851
+     "specificity": 0.7851,
+     "brier": 0.1529,
+     "ece": 0.126
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -1602,7 +1882,9 @@
      "f1": 0.7701,
      "precision": 0.7724,
      "recall": 0.7725,
-     "specificity": 0.7786
+     "specificity": 0.7786,
+     "brier": 0.1621,
+     "ece": 0.0924
     },
     {
      "species": "Soybean (Glycine max)",
@@ -1614,7 +1896,9 @@
      "f1": 0.7155,
      "precision": 0.7011,
      "recall": 0.7971,
-     "specificity": 0.7447
+     "specificity": 0.7447,
+     "brier": 0.1734,
+     "ece": 0.1699
     },
     {
      "species": "wheat",
@@ -1626,7 +1910,9 @@
      "f1": 0.7509,
      "precision": 0.7056,
      "recall": 0.8134,
-     "specificity": 0.6371
+     "specificity": 0.6371,
+     "brier": 0.1854,
+     "ece": 0.1087
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -1638,7 +1924,9 @@
      "f1": 0.7386,
      "precision": 0.7023,
      "recall": 0.8383,
-     "specificity": 0.3918
+     "specificity": 0.3918,
+     "brier": 0.2309,
+     "ece": 0.2759
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -1650,7 +1938,9 @@
      "f1": 0.6622,
      "precision": 0.7013,
      "recall": 0.6863,
-     "specificity": 0.6165
+     "specificity": 0.6165,
+     "brier": 0.2177,
+     "ece": 0.2235
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -1662,7 +1952,9 @@
      "f1": 0.6593,
      "precision": 0.6537,
      "recall": 0.7253,
-     "specificity": 0.4949
+     "specificity": 0.4949,
+     "brier": 0.2254,
+     "ece": 0.1924
     }
    ],
    "partial": false
@@ -1672,7 +1964,7 @@
    "classifier": "rf",
    "featureSet": "phys",
    "label": "Random forest \u00b7 Physicochemical",
-   "description": "Random forest \u2014 bagged decision trees; robust but usually a step behind XGBoost. 14 physicochemical descriptors only (length, charge, hydrophobicity, \u2026). Fast, no embeddings.",
+   "description": "Bagged decision trees. Robust, usually a step behind XGBoost. 14 physicochemical descriptors only (length, charge, hydrophobicity, \u2026). Fast, no embeddings.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -1739,6 +2031,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "Spirulina (Arthrospira platensis)",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.181,
+     "min": 0.1472,
+     "max": 0.204,
+     "minSpecies": "Spirulina (Arthrospira platensis)",
+     "maxSpecies": "Red rice bran (Oryza sativa)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1465,
+     "min": 0.0619,
+     "max": 0.2681,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -1752,7 +2060,9 @@
      "f1": 0.788,
      "precision": 0.7707,
      "recall": 0.826,
-     "specificity": 0.7283
+     "specificity": 0.7283,
+     "brier": 0.1515,
+     "ece": 0.1121
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -1764,7 +2074,9 @@
      "f1": 0.8018,
      "precision": 0.7896,
      "recall": 0.8239,
-     "specificity": 0.7706
+     "specificity": 0.7706,
+     "brier": 0.1472,
+     "ece": 0.0989
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -1776,7 +2088,9 @@
      "f1": 0.7579,
      "precision": 0.75,
      "recall": 0.7701,
-     "specificity": 0.7487
+     "specificity": 0.7487,
+     "brier": 0.1648,
+     "ece": 0.0619
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -1788,7 +2102,9 @@
      "f1": 0.749,
      "precision": 0.728,
      "recall": 0.8072,
-     "specificity": 0.5762
+     "specificity": 0.5762,
+     "brier": 0.192,
+     "ece": 0.2681
     },
     {
      "species": "Soybean (Glycine max)",
@@ -1800,7 +2116,9 @@
      "f1": 0.646,
      "precision": 0.659,
      "recall": 0.697,
-     "specificity": 0.7169
+     "specificity": 0.7169,
+     "brier": 0.1924,
+     "ece": 0.1645
     },
     {
      "species": "wheat",
@@ -1812,7 +2130,9 @@
      "f1": 0.7237,
      "precision": 0.6861,
      "recall": 0.7745,
-     "specificity": 0.6242
+     "specificity": 0.6242,
+     "brier": 0.1934,
+     "ece": 0.0886
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -1824,7 +2144,9 @@
      "f1": 0.6949,
      "precision": 0.7324,
      "recall": 0.6974,
-     "specificity": 0.6867
+     "specificity": 0.6867,
+     "brier": 0.2025,
+     "ece": 0.207
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -1836,7 +2158,9 @@
      "f1": 0.6967,
      "precision": 0.7086,
      "recall": 0.7247,
-     "specificity": 0.6185
+     "specificity": 0.6185,
+     "brier": 0.204,
+     "ece": 0.1712
     }
    ],
    "partial": false
@@ -1846,7 +2170,7 @@
    "classifier": "xgb",
    "featureSet": "phys",
    "label": "XGBoost \u00b7 Physicochemical",
-   "description": "Gradient-boosted decision trees (XGBoost) \u2014 the most consistent winner across datasets. 14 physicochemical descriptors only (length, charge, hydrophobicity, \u2026). Fast, no embeddings.",
+   "description": "Gradient-boosted decision trees. The most consistent winner across datasets. 14 physicochemical descriptors only (length, charge, hydrophobicity, \u2026). Fast, no embeddings.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -1913,6 +2237,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "Spirulina (Arthrospira platensis)",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.2008,
+     "min": 0.1622,
+     "max": 0.2296,
+     "minSpecies": "Spirulina (Arthrospira platensis)",
+     "maxSpecies": "Soybean (Glycine max)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1679,
+     "min": 0.0653,
+     "max": 0.2183,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Red rice bran (Oryza sativa)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -1926,7 +2266,9 @@
      "f1": 0.8122,
      "precision": 0.7824,
      "recall": 0.8552,
-     "specificity": 0.7457
+     "specificity": 0.7457,
+     "brier": 0.1685,
+     "ece": 0.1553
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -1938,7 +2280,9 @@
      "f1": 0.7966,
      "precision": 0.7895,
      "recall": 0.8119,
-     "specificity": 0.7723
+     "specificity": 0.7723,
+     "brier": 0.1622,
+     "ece": 0.1309
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -1950,7 +2294,9 @@
      "f1": 0.7616,
      "precision": 0.7475,
      "recall": 0.7794,
-     "specificity": 0.7417
+     "specificity": 0.7417,
+     "brier": 0.1657,
+     "ece": 0.0653
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -1962,7 +2308,9 @@
      "f1": 0.8344,
      "precision": 0.8231,
      "recall": 0.8792,
-     "specificity": 0.5668
+     "specificity": 0.5668,
+     "brier": 0.2292,
+     "ece": 0.2118
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -1974,7 +2322,9 @@
      "f1": 0.7017,
      "precision": 0.7197,
      "recall": 0.7183,
-     "specificity": 0.6611
+     "specificity": 0.6611,
+     "brier": 0.2189,
+     "ece": 0.2164
     },
     {
      "species": "Soybean (Glycine max)",
@@ -1986,7 +2336,9 @@
      "f1": 0.649,
      "precision": 0.6453,
      "recall": 0.7208,
-     "specificity": 0.6825
+     "specificity": 0.6825,
+     "brier": 0.2296,
+     "ece": 0.2106
     },
     {
      "species": "wheat",
@@ -1998,7 +2350,9 @@
      "f1": 0.7217,
      "precision": 0.6847,
      "recall": 0.7696,
-     "specificity": 0.6256
+     "specificity": 0.6256,
+     "brier": 0.2107,
+     "ece": 0.1344
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -2010,7 +2364,9 @@
      "f1": 0.7092,
      "precision": 0.7075,
      "recall": 0.7422,
-     "specificity": 0.6158
+     "specificity": 0.6158,
+     "brier": 0.2214,
+     "ece": 0.2183
     }
    ],
    "partial": false
@@ -2020,7 +2376,7 @@
    "classifier": "xgb",
    "featureSet": "pepbert",
    "label": "XGBoost \u00b7 PepBERT",
-   "description": "Gradient-boosted decision trees (XGBoost) \u2014 the most consistent winner across datasets. PepBERT protein-language-model embeddings of the peptide sequence.",
+   "description": "Gradient-boosted decision trees. The most consistent winner across datasets. PepBERT protein-language-model embeddings of the peptide sequence.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -2087,6 +2443,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1983,
+     "min": 0.1212,
+     "max": 0.3027,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1626,
+     "min": 0.063,
+     "max": 0.2647,
+     "minSpecies": "Potato (Solanum tuberosum)",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -2100,7 +2472,9 @@
      "f1": 0.8353,
      "precision": 0.8244,
      "recall": 0.8497,
-     "specificity": 0.7981
+     "specificity": 0.7981,
+     "brier": 0.1212,
+     "ece": 0.1036
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -2112,7 +2486,9 @@
      "f1": 0.8036,
      "precision": 0.7972,
      "recall": 0.8186,
-     "specificity": 0.7851
+     "specificity": 0.7851,
+     "brier": 0.1504,
+     "ece": 0.1079
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -2124,7 +2500,9 @@
      "f1": 0.7806,
      "precision": 0.7771,
      "recall": 0.7871,
-     "specificity": 0.7799
+     "specificity": 0.7799,
+     "brier": 0.1527,
+     "ece": 0.063
     },
     {
      "species": "Soybean (Glycine max)",
@@ -2136,7 +2514,9 @@
      "f1": 0.7229,
      "precision": 0.695,
      "recall": 0.8075,
-     "specificity": 0.7324
+     "specificity": 0.7324,
+     "brier": 0.1895,
+     "ece": 0.166
     },
     {
      "species": "wheat",
@@ -2148,7 +2528,9 @@
      "f1": 0.7588,
      "precision": 0.7197,
      "recall": 0.8101,
-     "specificity": 0.664
+     "specificity": 0.664,
+     "brier": 0.1858,
+     "ece": 0.1187
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -2160,7 +2542,9 @@
      "f1": 0.6984,
      "precision": 0.6867,
      "recall": 0.7438,
-     "specificity": 0.5761
+     "specificity": 0.5761,
+     "brier": 0.2397,
+     "ece": 0.2443
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -2172,7 +2556,9 @@
      "f1": 0.6924,
      "precision": 0.6974,
      "recall": 0.7256,
-     "specificity": 0.6102
+     "specificity": 0.6102,
+     "brier": 0.2442,
+     "ece": 0.2322
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -2184,7 +2570,9 @@
      "f1": 0.8102,
      "precision": 0.7944,
      "recall": 0.8821,
-     "specificity": 0.41
+     "specificity": 0.41,
+     "brier": 0.3027,
+     "ece": 0.2647
     }
    ],
    "partial": false
@@ -2261,6 +2649,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.2079,
+     "min": 0.1282,
+     "max": 0.3106,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1926,
+     "min": 0.0947,
+     "max": 0.3488,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -2274,7 +2678,9 @@
      "f1": 0.8133,
      "precision": 0.7946,
      "recall": 0.8374,
-     "specificity": 0.7739
+     "specificity": 0.7739,
+     "brier": 0.1282,
+     "ece": 0.0947
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -2286,7 +2692,9 @@
      "f1": 0.8084,
      "precision": 0.7936,
      "recall": 0.8329,
-     "specificity": 0.7732
+     "specificity": 0.7732,
+     "brier": 0.1579,
+     "ece": 0.1331
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -2298,7 +2706,9 @@
      "f1": 0.7783,
      "precision": 0.7728,
      "recall": 0.7872,
-     "specificity": 0.7734
+     "specificity": 0.7734,
+     "brier": 0.1857,
+     "ece": 0.1601
     },
     {
      "species": "Soybean (Glycine max)",
@@ -2310,7 +2720,9 @@
      "f1": 0.7185,
      "precision": 0.7165,
      "recall": 0.7735,
-     "specificity": 0.7616
+     "specificity": 0.7616,
+     "brier": 0.1854,
+     "ece": 0.179
     },
     {
      "species": "wheat",
@@ -2322,7 +2734,9 @@
      "f1": 0.7544,
      "precision": 0.7211,
      "recall": 0.7991,
-     "specificity": 0.6735
+     "specificity": 0.6735,
+     "brier": 0.2075,
+     "ece": 0.1708
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -2334,7 +2748,9 @@
      "f1": 0.625,
      "precision": 0.6004,
      "recall": 0.7079,
-     "specificity": 0.4211
+     "specificity": 0.4211,
+     "brier": 0.3106,
+     "ece": 0.3488
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -2346,7 +2762,9 @@
      "f1": 0.6815,
      "precision": 0.6645,
      "recall": 0.7515,
-     "specificity": 0.5091
+     "specificity": 0.5091,
+     "brier": 0.2476,
+     "ece": 0.2472
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -2358,7 +2776,9 @@
      "f1": 0.6919,
      "precision": 0.6807,
      "recall": 0.746,
-     "specificity": 0.5542
+     "specificity": 0.5542,
+     "brier": 0.2401,
+     "ece": 0.207
     }
    ],
    "partial": false
@@ -2368,7 +2788,7 @@
    "classifier": "rf",
    "featureSet": "pepbert",
    "label": "Random forest \u00b7 PepBERT",
-   "description": "Random forest \u2014 bagged decision trees; robust but usually a step behind XGBoost. PepBERT protein-language-model embeddings of the peptide sequence.",
+   "description": "Bagged decision trees. Robust, usually a step behind XGBoost. PepBERT protein-language-model embeddings of the peptide sequence.",
    "nSpecies": 8,
    "nDatasets": 30,
    "agg": {
@@ -2435,6 +2855,22 @@
      "minSpecies": "Faba bean (Vicia faba)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.1961,
+     "min": 0.1382,
+     "max": 0.2547,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1858,
+     "min": 0.1194,
+     "max": 0.3196,
+     "minSpecies": "wheat",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -2448,7 +2884,9 @@
      "f1": 0.8302,
      "precision": 0.8346,
      "recall": 0.8286,
-     "specificity": 0.8219
+     "specificity": 0.8219,
+     "brier": 0.1382,
+     "ece": 0.1551
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -2460,7 +2898,9 @@
      "f1": 0.7834,
      "precision": 0.7843,
      "recall": 0.7963,
-     "specificity": 0.7724
+     "specificity": 0.7724,
+     "brier": 0.1722,
+     "ece": 0.1562
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -2472,7 +2912,9 @@
      "f1": 0.7599,
      "precision": 0.7692,
      "recall": 0.7571,
-     "specificity": 0.7791
+     "specificity": 0.7791,
+     "brier": 0.1757,
+     "ece": 0.1261
     },
     {
      "species": "Soybean (Glycine max)",
@@ -2484,7 +2926,9 @@
      "f1": 0.7028,
      "precision": 0.6751,
      "recall": 0.7892,
-     "specificity": 0.71
+     "specificity": 0.71,
+     "brier": 0.1823,
+     "ece": 0.1758
     },
     {
      "species": "wheat",
@@ -2496,7 +2940,9 @@
      "f1": 0.7402,
      "precision": 0.7024,
      "recall": 0.7949,
-     "specificity": 0.6406
+     "specificity": 0.6406,
+     "brier": 0.195,
+     "ece": 0.1194
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -2508,7 +2954,9 @@
      "f1": 0.6816,
      "precision": 0.6801,
      "recall": 0.7353,
-     "specificity": 0.555
+     "specificity": 0.555,
+     "brier": 0.2175,
+     "ece": 0.2014
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -2520,7 +2968,9 @@
      "f1": 0.6469,
      "precision": 0.6841,
      "recall": 0.6966,
-     "specificity": 0.5675
+     "specificity": 0.5675,
+     "brier": 0.2332,
+     "ece": 0.2331
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -2532,7 +2982,9 @@
      "f1": 0.7049,
      "precision": 0.6921,
      "recall": 0.7872,
-     "specificity": 0.3767
+     "specificity": 0.3767,
+     "brier": 0.2547,
+     "ece": 0.3196
     }
    ],
    "partial": false
@@ -2609,6 +3061,22 @@
      "minSpecies": "Red rice bran (Oryza sativa)",
      "maxSpecies": "gracilaria",
      "nSpecies": 8
+    },
+    "brier": {
+     "mean": 0.2003,
+     "min": 0.1378,
+     "max": 0.2363,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Red rice bran (Oryza sativa)",
+     "nSpecies": 8
+    },
+    "ece": {
+     "mean": 0.1695,
+     "min": 0.1039,
+     "max": 0.2554,
+     "minSpecies": "gracilaria",
+     "maxSpecies": "Faba bean (Vicia faba)",
+     "nSpecies": 8
     }
    },
    "bySpecies": [
@@ -2622,7 +3090,9 @@
      "f1": 0.8234,
      "precision": 0.7979,
      "recall": 0.8571,
-     "specificity": 0.7733
+     "specificity": 0.7733,
+     "brier": 0.1378,
+     "ece": 0.1039
     },
     {
      "species": "Spirulina (Arthrospira platensis)",
@@ -2634,7 +3104,9 @@
      "f1": 0.7919,
      "precision": 0.7815,
      "recall": 0.8151,
-     "specificity": 0.7621
+     "specificity": 0.7621,
+     "brier": 0.1581,
+     "ece": 0.1185
     },
     {
      "species": "Faba bean (Vicia faba)",
@@ -2646,7 +3118,9 @@
      "f1": 0.741,
      "precision": 0.7295,
      "recall": 0.7976,
-     "specificity": 0.5743
+     "specificity": 0.5743,
+     "brier": 0.2202,
+     "ece": 0.2554
     },
     {
      "species": "Potato (Solanum tuberosum)",
@@ -2658,7 +3132,9 @@
      "f1": 0.7467,
      "precision": 0.7403,
      "recall": 0.7579,
-     "specificity": 0.7378
+     "specificity": 0.7378,
+     "brier": 0.1863,
+     "ece": 0.1183
     },
     {
      "species": "Soybean (Glycine max)",
@@ -2670,7 +3146,9 @@
      "f1": 0.6638,
      "precision": 0.6707,
      "recall": 0.7191,
-     "specificity": 0.73
+     "specificity": 0.73,
+     "brier": 0.2327,
+     "ece": 0.2195
     },
     {
      "species": "wheat",
@@ -2682,7 +3160,9 @@
      "f1": 0.7156,
      "precision": 0.6822,
      "recall": 0.7612,
-     "specificity": 0.6242
+     "specificity": 0.6242,
+     "brier": 0.2193,
+     "ece": 0.1475
     },
     {
      "species": "Tetradesmus obliquus (microalga)",
@@ -2694,7 +3174,9 @@
      "f1": 0.7004,
      "precision": 0.6996,
      "recall": 0.7491,
-     "specificity": 0.6011
+     "specificity": 0.6011,
+     "brier": 0.2119,
+     "ece": 0.1918
     },
     {
      "species": "Red rice bran (Oryza sativa)",
@@ -2706,7 +3188,9 @@
      "f1": 0.6507,
      "precision": 0.6505,
      "recall": 0.7177,
-     "specificity": 0.4784
+     "specificity": 0.4784,
+     "brier": 0.2363,
+     "ece": 0.2012
     }
    ],
    "partial": false

website/src/data/mockData.ts

@@ -179,7 +179,7 @@ export const CLASSIFIER_OPTIONS = [
 ]
 
 export const SPLIT_OPTIONS = [
-  { id: 'protein', label: 'Group by protein', desc: 'Production default — avoids peptide leakage' },
+  { id: 'protein', label: 'Group by protein', desc: 'Production default, avoids peptide leakage' },
   { id: 'cluster90', label: 'Cluster (90% id)', desc: 'MMseqs2 sequence-identity clusters' },
   { id: 'cluster50', label: 'Cluster (50% id)', desc: 'Stricter homology grouping' },
   { id: 'peptide', label: 'By peptide', desc: 'Leakage baseline (inflates AUROC)' },

website/src/pages/Benchmarking.tsx

@@ -13,7 +13,8 @@ import stats from '../data/generated/cross_model_stats.json'
 const fmt = (v: number | null | undefined) => (v == null ? '—' : v.toFixed(3))
 
 // Metric columns shown in the model-type leaderboard (order matters).
-const METRIC_COLS = ['auprc', 'accuracy', 'mcc', 'f1', 'recall', 'specificity', 'precision'] as const
+// brier + ece are calibration metrics where LOWER is better.
+const METRIC_COLS = ['auprc', 'accuracy', 'mcc', 'f1', 'recall', 'specificity', 'precision', 'brier', 'ece'] as const
 
 // ─────────────────────────────────────────────────────────────── Model-type leaderboard
 function TypeLeaderboard() {
@@ -41,16 +42,15 @@ function TypeLeaderboard() {
   return (
     <div>
       <p className="muted" style={{ marginTop: 0 }}>
-        Every classifier × feature-set combination, evaluated on all {LEADERBOARD.nDatasets} datasets
-        (5-fold protein-grouped CV) and aggregated across the {LEADERBOARD.nSpecies} species. The headline
-        is the <strong>mean AUROC across species</strong>; the range shows the weakest → strongest species.
-        Click a row for its description and per-species breakdown.
+        Every classifier and feature-set combo, scored on all {LEADERBOARD.nDatasets} datasets and averaged
+        across {LEADERBOARD.nSpecies} species. Sorted by <strong>mean AUROC</strong>; the range shows
+        weakest to strongest species. Click a row for details.
       </p>
 
       <div className="stat-row">
         <div className="stat">
           <div className="stat-value">{best.agg.auroc?.mean.toFixed(3)}</div>
-          <div className="stat-label">Best mean AUROC — {best.label}</div>
+          <div className="stat-label">Top model: {best.label}</div>
         </div>
         <div className="stat">
           <div className="stat-value">{LEADERBOARD.nSpecies}</div>
@@ -78,6 +78,7 @@ function TypeLeaderboard() {
               {METRIC_COLS.map((k) => (
                 <th key={k} className="num">
                   {labels[k]}
+                  {(k === 'brier' || k === 'ece') && ' ↓'}
                 </th>
               ))}
             </tr>
@@ -154,6 +155,8 @@ function TypeDetail({ m }: { m: ModelType }) {
                 <th className="num">F1</th>
                 <th className="num">Sens.</th>
                 <th className="num">Spec.</th>
+                <th className="num">Brier ↓</th>
+                <th className="num">ECE ↓</th>
               </tr>
             </thead>
             <tbody>
@@ -166,6 +169,8 @@ function TypeDetail({ m }: { m: ModelType }) {
                   <td className="num">{fmt(s.f1 as number)}</td>
                   <td className="num">{fmt(s.recall as number)}</td>
                   <td className="num">{fmt(s.specificity as number)}</td>
+                  <td className="num">{fmt(s.brier as number)}</td>
+                  <td className="num">{fmt(s.ece as number)}</td>
                 </tr>
               ))}
             </tbody>
@@ -187,8 +192,8 @@ function IndividualModels() {
   return (
     <div>
       <p className="muted" style={{ marginTop: 0 }}>
-        The {MODELS.length} deployed per-dataset models — each is the winning classifier × feature-set for
-        that specific hydrolysate. Click for the paper, configuration, and caveats.
+        The {MODELS.length} deployed models, one per dataset. Each is the winning classifier and feature
+        set for that hydrolysate. Click for the paper, config, and caveats.
       </p>
       <div className="table-wrap">
         <table className="table">
@@ -275,15 +280,14 @@ function Analysis() {
   return (
     <div>
       <p className="muted" style={{ marginTop: 0 }}>
-        Cross-dataset statistics over {stats.nContexts} contexts with full ablations (Friedman tests on
-        mean CV ranks). Higher mean rank = better; charts show every treatment on the same scale.
+        What drives detectability, across {stats.nContexts} datasets. Higher mean rank is better.
       </p>
 
       <div className="grid-2">
-        <Card title="Feature set" subtitle={`What to represent the peptide with · Friedman p = ${fs.friedmanP?.toExponential(1)}`}>
+        <Card title="Feature set" subtitle={`How to represent the peptide · Friedman p = ${fs.friedmanP?.toExponential(1)}`}>
           <RankBars ranks={fs.meanRanks} labelFn={featureSetLabel} best={fs.bestByRank} />
           <p className="muted" style={{ fontSize: '0.82rem', marginBottom: 0, marginTop: 10 }}>
-            Embeddings + physicochemical descriptors win; physicochemical alone is weakest.
+            Embeddings plus physicochemical features win. Physicochemical alone is weakest.
           </p>
         </Card>
 
@@ -295,7 +299,7 @@ function Analysis() {
         </Card>
       </div>
 
-      <Card title="Split rigor — the leakage check" subtitle="Mean AUROC by how train/test are separated; whiskers are the 95% CI">
+      <Card title="Split rigor (leakage check)" subtitle="Mean AUROC by how train and test are separated. Whiskers show the 95% CI.">
         <div style={{ maxWidth: 560 }}>
           {tiers.map((t) => {
             const s = split[t.key]
@@ -319,8 +323,8 @@ function Analysis() {
           })}
         </div>
         <p className="muted" style={{ fontSize: '0.82rem', marginBottom: 0, marginTop: 10 }}>
-          Peptide-level splitting inflates AUROC — the same protein leaks across train/test. HydroPD trains
-          on the <strong>protein split</strong> (production), so reported numbers aren't leakage-inflated.
+          Splitting by peptide inflates AUROC, because the same protein leaks across train and test.
+          HydroPD trains on the <strong>protein split</strong>, so these numbers aren't inflated.
         </p>
       </Card>
 
@@ -328,9 +332,10 @@ function Analysis() {
         <p style={{ marginBottom: 6 }}>{neg.method}</p>
         {neg.seed_spread_ci && (
           <p className="muted" style={{ margin: 0 }}>
-            AUROC moves only <strong>{neg.seed_spread_ci.mean.toFixed(2)}</strong> (95% CI{' '}
-            {neg.seed_spread_ci.lo.toFixed(2)}–{neg.seed_spread_ci.hi.toFixed(2)}) across negative-sampling
-            seeds — the choice of random negatives barely affects performance.
+            Changing the negative-sampling seed moves AUROC by only{' '}
+            <strong>{neg.seed_spread_ci.mean.toFixed(2)}</strong> (95% CI{' '}
+            {neg.seed_spread_ci.lo.toFixed(2)}–{neg.seed_spread_ci.hi.toFixed(2)}). The choice of random
+            negatives barely matters.
           </p>
         )}
       </Card>
@@ -347,8 +352,8 @@ export default function Benchmarking() {
         <div className="eyebrow">Leaderboards</div>
         <h1>Benchmarking</h1>
         <p className="lead">
-          How the model types compare across species, the {MODELS.length} deployed per-dataset models, and
-          the cross-model analysis of what actually drives detectability.
+          Compare model types, browse the {MODELS.length} deployed models, and see what drives
+          detectability.
         </p>
       </div>
 

website/src/pages/BioactivityScreening.tsx

@@ -72,7 +72,7 @@ const PEPTIDE_COLUMNS: Column<PeptideView>[] = [
     sortable: true,
     render: (r) =>
       r.igeContains ? (
-        <span className="badge badge-warn" title="The peptide equals or fully contains a documented linear IgE epitope (IEDB) — a potential allergenicity concern.">
+        <span className="badge badge-warn" title="Holds a documented linear IgE epitope (IEDB). Potential allergenicity concern.">
           contains
         </span>
       ) : (
@@ -106,7 +106,7 @@ function LiveBadge({ live }: { live: boolean }) {
       title={
         live
           ? 'Matched against the HydroPD bioactivity/allergen master database (server-side).'
-          : 'Backend unavailable — screening needs the server; no results computed offline.'
+          : 'Backend unavailable. Screening runs on the server.'
       }
     >
       {live ? 'Live screening' : 'Offline (backend unavailable)'}
@@ -196,17 +196,15 @@ export default function BioactivityScreening() {
         <div className="eyebrow">Tool</div>
         <h1>Bioactivity &amp; Allergen Screening</h1>
         <p className="lead">
-          Screen sequences against HydroPD's centralized master database (10 bioactivity
-          databases, ~58k peptides, plus IgE epitopes and recognized-allergen proteins).
-          Two match types: <strong>bioactivity, cytotoxicity, and allergen origin are
-          exact whole-sequence matches</strong> (I/L-folded), while the{' '}
-          <strong>IgE-epitope check is a “contains” match</strong> — the peptide equals or
-          fully contains a documented linear epitope.
+          Check sequences against 10 bioactivity databases (~58k peptides) plus IgE epitopes
+          and recognized allergens. Bioactivity, cytotoxicity, and allergen origin use{' '}
+          <strong>exact matches</strong>. The IgE check uses a <strong>contains match</strong>
+          {' '}(the peptide holds a known epitope).
         </p>
       </div>
 
       {/* Peptide bioactivity + IgE (Q2) */}
-      <Card title="1 · Screen peptides — bioactivity, cytotoxicity &amp; IgE epitopes">
+      <Card title="1 · Peptides: bioactivity, cytotoxicity & IgE">
         <SequenceInput
           label="Peptide sequences"
           value={pepInput}
@@ -287,11 +285,11 @@ export default function BioactivityScreening() {
       </Card>
 
       {/* Protein allergen origin (Q1) */}
-      <Card title="2 · Screen proteins — allergen origin">
+      <Card title="2 · Proteins: allergen origin">
         <p className="muted" style={{ marginTop: 0, fontSize: '0.86rem' }}>
-          A bare peptide can't answer this — allergen origin is a protein-level question.
-          Paste full protein sequences (or FASTA) to check for an <strong>exact
-          whole-protein match</strong> against recognized allergens (WHO/IUIS ∪ AllergenOnline).
+          Allergen origin is a protein-level question, so paste full protein sequences (or FASTA).
+          We check for an <strong>exact match</strong> against recognized allergens (WHO/IUIS and
+          AllergenOnline).
         </p>
         <SequenceInput
           label="Protein sequences"

website/src/pages/Data.tsx

@@ -57,7 +57,7 @@ function Computational() {
     <div>
       <Card
         title="Download datasets"
-        subtitle="Raw peptidomics tables + the literature workbook. Everything here reflects the current HydroPD revamp (Steps 0–6)."
+        subtitle="Raw peptidomics tables and the literature workbook."
       >
         <DownloadCard
           title="Raw data bundle (H1–H13)"
@@ -187,7 +187,7 @@ function Experimentalist() {
     <div>
       <Card
         title="What data to generate & retain"
-        subtitle="Producing compositional / peptidomics data? This is the HydroPD reporting schema — the fields to capture, in two tiers: MINIMUM (enough to interpret + model) and IDEAL (report everything)."
+        subtitle="Making peptidomics data? Capture these fields. Two tiers: MINIMUM to model, IDEAL to report everything."
         actions={
           <a
             className="btn btn-secondary btn-sm"
@@ -243,7 +243,7 @@ function Experimentalist() {
 
       <Card
         title="Where new data is most needed"
-        subtitle="High-value raw materials for cultivated-meat media that still lack usable peptidomics data — the best places to generate it."
+        subtitle="Raw materials that lack usable peptidomics data. The best places to generate it."
         actions={
           <div className="segmented" style={{ padding: 2 }}>
             <button
@@ -263,10 +263,8 @@ function Experimentalist() {
           </div>
         }
       >
-        <p style={{ marginTop: 0 }}>
-          High-value raw materials for cultivated-meat media that lack usable
-          peptidomics data — the most useful places to generate new data. Sorted by{' '}
-          <strong>{gapSort === 'papers' ? 'fewest papers found' : 'CCM priority'}</strong>.
+        <p className="muted" style={{ marginTop: 0, fontSize: '0.84rem' }}>
+          Sorted by <strong>{gapSort === 'papers' ? 'fewest papers found' : 'CCM priority'}</strong>.
         </p>
         <div className="table-wrap">
           <table className="table">
@@ -310,10 +308,8 @@ function Experimentalist() {
           </table>
         </div>
         <p className="muted" style={{ fontSize: '0.78rem', marginTop: 10, marginBottom: 0 }}>
-          {dataGaps.rawMaterials.filter((r) => r.status === 'none').length} raw
-          materials returned <strong>zero papers</strong> in the literature search
-          (e.g. Canola is <em>not</em> among them — Brassica napus has 1 found paper
-          but no usable data yet).
+          {dataGaps.rawMaterials.filter((r) => r.status === 'none').length} raw materials returned{' '}
+          <strong>zero papers</strong> in the search.
         </p>
       </Card>
 
@@ -394,8 +390,8 @@ export default function Data() {
         <div className="eyebrow">Data</div>
         <h1>Data</h1>
         <p className="lead">
-          Download the datasets behind HydroPD, review the literature-search
-          landscape, and see which raw-material data gaps are most worth filling.
+          Download the datasets, browse the literature landscape, and see which data
+          gaps are worth filling.
         </p>
       </div>
 

website/src/pages/DetectabilityPrediction.tsx

@@ -41,11 +41,11 @@ const COLUMNS: Column<DetectabilityRow>[] = [
     sortable: true,
     render: (r) =>
       r.inTraining === 'positive' ? (
-        <span className="badge badge-warn" title="This peptide was a detected (positive) example in this model's training set — the score may reflect memorisation, not generalisation.">
+        <span className="badge badge-warn" title="Detected (positive) example in this model's training set. The score may reflect memorisation.">
           Train⁺
         </span>
       ) : r.inTraining === 'negative' ? (
-        <span className="badge badge-warn" title="This peptide was an undetected (negative) example in this model's training set — the score may reflect memorisation, not generalisation.">
+        <span className="badge badge-warn" title="Undetected (negative) example in this model's training set. The score may reflect memorisation.">
           Train⁻
         </span>
       ) : (
@@ -86,10 +86,8 @@ export default function DetectabilityPrediction() {
         <div className="eyebrow">Tool</div>
         <h1>Detectability Prediction</h1>
         <p className="lead">
-          Paste peptide sequences and pick one of the 30 trained context-specific
-          models (species + enzyme). Describe your conditions to auto-rank the
-          best-matched model, and any peptide that was in the model's training set is
-          flagged.
+          Paste peptides and pick a model. Describe your conditions to auto-rank the
+          best fit, or choose one yourself. Peptides seen during training are flagged.
         </p>
       </div>
 
@@ -158,7 +156,7 @@ export default function DetectabilityPrediction() {
               title={
                 live
                   ? 'Scores computed by the model (server-side inference).'
-                  : 'Backend unavailable — showing illustrative offline scores.'
+                  : 'Backend unavailable. Showing illustrative offline scores.'
               }
             >
               {live ? 'Live inference' : 'Offline preview'}

website/src/pages/DetectabilityTraining.tsx

@@ -63,9 +63,8 @@ export default function DetectabilityTraining() {
         <div className="eyebrow">Tool</div>
         <h1>Train a Detectability Model</h1>
         <p className="lead">
-          Provide detected (positive) peptides and a species tax ID, choose model
-          specifications or the recommended defaults, and train a context-specific
-          detectability model following the HydroPD Step-4 protocol.
+          Give detected peptides and a species tax ID, pick your specs or use the
+          defaults, and train a model. Then download the model and processed data.
         </p>
       </div>
 
@@ -105,7 +104,7 @@ export default function DetectabilityTraining() {
             >
               {FEATURE_SET_OPTIONS.map((o) => (
                 <option key={o.id} value={o.id}>
-                  {o.label} — {o.desc}
+                  {o.label} · {o.desc}
                 </option>
               ))}
             </select>
@@ -117,7 +116,7 @@ export default function DetectabilityTraining() {
             >
               {CLASSIFIER_OPTIONS.map((o) => (
                 <option key={o.id} value={o.id}>
-                  {o.label} — {o.desc}
+                  {o.label} · {o.desc}
                 </option>
               ))}
             </select>
@@ -129,7 +128,7 @@ export default function DetectabilityTraining() {
             >
               {SPLIT_OPTIONS.map((o) => (
                 <option key={o.id} value={o.id}>
-                  {o.label} — {o.desc}
+                  {o.label} · {o.desc}
                 </option>
               ))}
             </select>
@@ -188,8 +187,8 @@ export default function DetectabilityTraining() {
           <>
             <p className="muted" style={{ fontSize: '0.8rem', marginTop: 0 }}>
               {live
-                ? '● Trained on the backend — real Step 2–5 pipeline (assign → k-mer negatives → protein-grouped CV).'
-                : '○ Offline preview (illustrative numbers) — the training backend is not reachable.'}
+                ? '● Trained on the server. Real pipeline: assign proteins, sample negatives, protein-grouped CV.'
+                : '○ Offline preview with illustrative numbers. The training backend is not reachable.'}
             </p>
             {errorMsg && (
               <p

website/src/pages/Home.tsx

@@ -5,25 +5,31 @@ const TOOLS = [
     to: '/detectability',
     icon: '📈',
     title: 'Detectability Prediction',
-    desc: 'Score peptides for MS detectability using one of 30 context-specific trained models.',
+    desc: 'Score your peptides against one of 30 species-specific models.',
+  },
+  {
+    to: '/bioactivity',
+    icon: '🧬',
+    title: 'Bioactivity Screening',
+    desc: 'Check peptides for known bioactivity, cytotoxicity, and allergen origin.',
   },
   {
     to: '/training',
     icon: '🧪',
     title: 'Train a Model',
-    desc: 'Train a detectability model from your own detected peptides and species.',
+    desc: 'Train a model on your own peptides and species, then download it.',
   },
   {
     to: '/benchmarking',
     icon: '🏆',
     title: 'Benchmarking',
-    desc: 'Leaderboard of the 30 trained models and cross-model analysis of what works.',
+    desc: 'Compare model types and see what drives performance.',
   },
   {
     to: '/data',
     icon: '🗄️',
     title: 'Data',
-    desc: 'Download the raw datasets, review the literature-search summary, and see data gaps to fill.',
+    desc: 'Download datasets, browse the literature landscape, and find gaps to fill.',
   },
 ]
 
@@ -36,9 +42,8 @@ export default function Home() {
         </div>
         <h1>HydroPD</h1>
         <p>
-          Machine-learning tools for mass-spectrometry detectability of peptides
-          from non-animal food-protein hydrolysates — built to support
-          cell-culture media optimization for cultivated meat.
+          Predict which peptides show up in LC-MS/MS. Built for non-animal protein
+          hydrolysates used in cultivated-meat media.
         </p>
       </div>