feat: add CatRange integration (Realkcat update)

Dockerfile.envs

@@ -116,6 +116,13 @@ RUN --mount=type=cache,target=/opt/conda/pkgs,sharing=locked \
     mamba create -n realkcat_env python=3.9 -c conda-forge -y \
     && conda run -n realkcat_env pip install -r docker-requirements/realkcat_requirements.txt
 
+FROM base AS env-catrange
+COPY docker-requirements/catrange_requirements.txt ./docker-requirements/
+RUN --mount=type=cache,target=/opt/conda/pkgs,sharing=locked \
+    --mount=type=cache,id=webkinpred-pip-py310,target=/root/.cache/pip,sharing=locked \
+    mamba create -n catrange_env python=3.10 -c conda-forge -y \
+    && conda run -n catrange_env pip install -r docker-requirements/catrange_requirements.txt
+
 FROM base AS env-pseq2sites
 RUN --mount=type=cache,target=/opt/conda/pkgs,sharing=locked \
     --mount=type=cache,id=webkinpred-pip-py37,target=/root/.cache/pip,sharing=locked \
@@ -161,6 +168,7 @@ COPY --from=env-mmisakm   /opt/conda/envs/mmisakm_env  /opt/conda/envs/mmisakm_e
 COPY --from=env-omniesi   /opt/conda/envs/omniesi_env  /opt/conda/envs/omniesi_env
 COPY --from=env-iecata    /opt/conda/envs/iecata_env   /opt/conda/envs/iecata_env
 COPY --from=env-realkcat   /opt/conda/envs/realkcat_env  /opt/conda/envs/realkcat_env
+COPY --from=env-catrange   /opt/conda/envs/catrange_env /opt/conda/envs/catrange_env
 COPY --from=env-pseq2sites /opt/conda/envs/pseq2sites    /opt/conda/envs/pseq2sites
 COPY --from=env-esm        /opt/conda/envs/esm           /opt/conda/envs/esm
 COPY --from=env-esmc       /opt/conda/envs/esmc          /opt/conda/envs/esmc

README.md

@@ -36,7 +36,7 @@ Open Kinetics Predictor is a production web interface for predicting enzyme kine
 | CataPro | Protein sequence + substrate SMILES | kcat, Km, or kcat/Km | [Wang et al., Nat Commun 2025](https://www.nature.com/articles/s41467-025-58038-4) ([GitHub](https://github.com/zchwang/CataPro)) |
 | CatPred | Protein sequence + substrate SMILES | kcat or Km | [Boorla et al., Nat Commun 2025](https://www.nature.com/articles/s41467-025-57215-9) ([GitHub](https://github.com/maranasgroup/CatPred)) |
 | OmniESI | Protein sequence + substrate SMILES | kcat or Km | [Nie et al., arXiv 2025](https://doi.org/10.48550/arXiv.2506.17963) ([GitHub](https://github.com/Hong-yu-Zhang/OmniESI)) |
-| RealKcat | Protein sequence + substrate SMILES | kcat or Km | [Sajeevan et al., bioRxiv 2025](https://doi.org/10.1101/2025.02.10.637555) ([GitHub](https://github.com/TKAI-LAB-Mali/RealKcat)) |
+| CatRange | Protein sequence + substrate SMILES | kcat or Km | [Sajeevan et al., bioRxiv 2025](https://doi.org/10.1101/2025.02.10.637555) ([GitHub](https://github.com/ssbio/CatRange)) |
 | IECata | Protein sequence + substrate SMILES | kcat/Km | [Wang et al., Brief Bioinform 2025](https://doi.org/10.1093/bib/bbaf283) ([GitHub](https://github.com/zhaoyanpeng208/IECata)) |
 | MMISA-KM | Protein sequence + substrate SMILES | Km | [Song & Wang, DDCLS 2025](https://doi.org/10.1109/DDCLS66240.2025.11064981) ([GitHub](https://github.com/kaiwang-group/MMISA-KM)) |
 

api/methods/catrange.py

@@ -0,0 +1,36 @@
+from api.methods.base import MethodDescriptor, SubprocessEngineConfig
+
+
+descriptor = MethodDescriptor(
+    key="CatRange",
+    display_name="CatRange",
+    authors=(
+        "Karuna Anna Sajeevan, Abraham Osinuga, Arunraj B, Sakib Ferdous, "
+        "Nabia Shahreen, Shashank Koneru, Laura Mariana Santos-Correa, "
+        "Rahil Salehi, Niaz Bahar Chowdhury, Randy Aryee, Brisa Calderon-Lopez, "
+        "Supantha Dey, Ankur Mali, Rajib Saha, and Ratul Chowdhury"
+    ),
+    publication_title="CatRange Enables Robust Prediction of Enzyme Variant Kinetic Regimes",
+    citation_url="https://doi.org/10.1101/2025.02.10.637555",
+    repo_url="https://github.com/ssbio/CatRange",
+    supports=["kcat", "Km"],
+    input_format="single",
+    output_cols={
+        "kcat": "Predicted kcat range: kcat (1/s)",
+        "Km": "Predicted KM range: kM (M)",
+    },
+    max_seq_len=1024,
+    col_to_kwarg={"Substrate": "substrates"},
+    target_kwargs={
+        "kcat": {"kinetics_type": "KCAT"},
+        "Km": {"kinetics_type": "KM"},
+    },
+    subprocess=SubprocessEngineConfig(
+        python_path_key="CatRange",
+        script_key="CatRange",
+        data_path_env={
+            "CATRANGE_REPO_ROOT": "CatRange",
+            "CATRANGE_MODELS_DIR": "CatRange",
+        },
+    ),
+)

api/methods/realkcat.py

@@ -1,32 +1 @@
-from api.methods.base import MethodDescriptor, SubprocessEngineConfig
-
-descriptor = MethodDescriptor(
-    key="RealKcat",
-    display_name="RealKcat",
-    authors="Sajeevan AK, Osinuga A, Mali A, et al.",
-    publication_title="Robust Prediction of Enzyme Variant Kinetics with RealKcat",
-    citation_url="https://doi.org/10.1101/2025.02.10.637555",
-    repo_url="https://github.com/TKAI-LAB-Mali/RealKcat",
-
-    supports=["kcat", "Km"],
-    input_format="single",
-    output_cols={"kcat": "kcat (1/s)", "Km": "Km (M)"},
-    max_seq_len=1022,  # ESM2 context limit
-
-    col_to_kwarg={"Substrate": "substrates"},
-    target_kwargs={
-        "kcat": {"kinetics_type": "KCAT"},
-        "Km": {"kinetics_type": "KM"}
-    },
-
-    subprocess=SubprocessEngineConfig(
-        python_path_key="RealKcat",
-        script_key="RealKcat",
-        data_path_env={
-            "REALKCAT_DATA": "RealKcat_DATA",
-            "REALKCAT_EMBED_CACHE_DIR": "realkcat_esm2_last_mean",
-        },
-    ),
-
-    embeddings_used=["realkcat_esm2_last_mean"],
-)
+"""Legacy RealKcat module retained as an empty stub after the method was replaced by CatRange."""

docker-requirements/catrange_requirements.txt

@@ -0,0 +1,12 @@
+numpy==1.26.4
+pandas==2.2.3
+tqdm==4.19.0
+joblib==1.4.2
+scikit-learn==1.1.3
+transformers==4.44.2
+tokenizers==0.19.1
+huggingface-hub==0.24.7
+safetensors==0.4.5
+# CatRange runtime uses the local ESM-C package from the repo checkout; install it from source.
+esm @ git+https://github.com/evolutionaryscale/esm.git
+torch==2.4.0

docs/index.rst

@@ -40,7 +40,7 @@ Open Kinetics Predictor
      <article class="feature-card">
        <p class="feature-kicker">Compare</p>
        <h2>Use published engines</h2>
-       <p>Access KinForm, UniKP, DLKcat, TurNup, EITLEM, CataPro, CatPred, OmniESI, RealKcat, IECata, and MMISA-KM through one interface.</p>
+       <p>Access KinForm, UniKP, DLKcat, TurNup, EITLEM, CataPro, CatPred, OmniESI, CatRange, IECata, and MMISA-KM through one interface.</p>
      </article>
      <article class="feature-card">
        <p class="feature-kicker">Extend</p>

docs/project/supported_methods.rst

@@ -23,7 +23,7 @@ Supported engines
 - CatPred, kcat and Km.
 - OmniESI, kcat and Km.
 - OmniESI + O2DENet, kcat and Km.
-- RealKcat, kcat and Km.
+- CatRange, kcat and Km.
 - IECata, kcat/Km.
 - MMISA-KM, Km.
 

frontend/src/components/ApiDocs/ApiDocs.jsx

@@ -574,7 +574,7 @@ export default function ApiDocs() {
     "KinForm-L": 0,
     "CatPred": 0,
     "OmniESI": 0,
-    "RealKcat": 0,
+    "CatRange": 0,
     "IECata": 1,
     "MMISA-KM": 0,
     "Server": 0
@@ -589,7 +589,7 @@ export default function ApiDocs() {
     "KinForm-L": 1500,
     "CatPred": 2048,
     "OmniESI": 1000,
-    "RealKcat": 1022,
+    "CatRange": 1024,
     "IECata": 1000,
     "MMISA-KM": 500,
     "Server": 10000
@@ -905,10 +905,10 @@ export default function ApiDocs() {
                 <td>1,000 residues</td>
               </tr>
               <tr>
-                <td><code>RealKcat</code></td>
+                <td><code>CatRange</code></td>
                 <td>kcat or Km</td>
                 <td><code>Protein Sequence</code>, <code>Substrate</code></td>
-                <td>1,022 residues</td>
+                <td>1,024 residues</td>
               </tr>
               <tr>
                 <td><code>IECata</code></td>

frontend/src/components/job-submission/components/HowToUseCard.jsx

@@ -10,7 +10,7 @@ const FORMAT_SPECS = [
   {
     key: 'single',
     name: 'Single Substrate',
-    methods: ['CataPro', 'CatPred (Km)', 'DLKcat', 'EITLEM', 'IECata', 'KinForm-H', 'KinForm-L', 'MMISA-KM', 'OmniESI', 'RealKcat', 'UniKP'],
+    methods: ['CataPro', 'CatPred (Km)', 'CatRange', 'DLKcat', 'EITLEM', 'IECata', 'KinForm-H', 'KinForm-L', 'MMISA-KM', 'OmniESI', 'UniKP'],
     columns: [
       { col: 'Protein Sequence', desc: 'full amino-acid sequence' },
       { col: 'Substrate',        desc: 'SMILES or InChI — one per row' },

models/CatRange/.gitignore

@@ -0,0 +1,37 @@
+__pycache__/
+*.py[cod]
+.ipynb_checkpoints/
+.DS_Store
+.pytest_cache/
+*.log
+.venv/
+envs/*/
+
+# CatRange XGBoost model binaries are hundreds of MB and should be released
+# through Zenodo/Hugging Face/GitHub Releases or another model host.
+inference/models/*.pkl
+inference/models/*fold*.pt
+catrange_model/outputs/
+catrange_model/logs/
+catrange_model/data/processed/
+catrange_model/data_robust_v1/
+catrange_model/data_robust_v1
+reproducibility_runs/
+
+# External benchmark caches/checkpoints are generated, very large, and
+# reproducible from the scripts plus manifests.
+benchmarks/retrained_comparators/**/catpred_esm2_cache/
+benchmarks/retrained_comparators/**/checkpoints/
+benchmarks/retrained_comparators/**/*.pt
+benchmarks/retrained_comparators/**/*.pth
+benchmarks/retrained_comparators/**/*.pkl
+benchmarks/retrained_comparators/**/*.joblib
+benchmarks/retrained_comparators/tmp*/
+benchmarks/retrained_comparators/vendor/
+
+# Local model/package build products.
+*.egg-info/
+build/
+dist/
+
+ignore/