Add CatPred integration scaffold

Author: theproteinbot

Dockerfile

@@ -106,6 +106,15 @@ RUN --mount=type=cache,target=/opt/conda/pkgs,sharing=locked \
     && mamba install -n catapro_env -c conda-forge rdkit=2024.03.6 -y \
     && conda run -n catapro_env pip install -r docker-requirements/catapro_requirements.txt
 
+# ── CatPred ───────────────────────────────────────────────────────────────────
+FROM base AS env-catpred
+COPY docker-requirements/catpred_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 catpred_env python=3.10.15 -c conda-forge -y \
+    && mamba install -n catpred_env -c conda-forge rdkit=2024.03.6 -y \
+    && conda run -n catpred_env pip install -r docker-requirements/catpred_requirements.txt
+
 # ── pseq2sites ────────────────────────────────────────────────────────────────
 FROM base AS env-pseq2sites
 RUN --mount=type=cache,target=/opt/conda/pkgs,sharing=locked \
@@ -155,6 +164,7 @@ COPY --from=env-eitlem    /opt/conda/envs/eitlem_env   /opt/conda/envs/eitlem_en
 COPY --from=env-turnup    /opt/conda/envs/turnup_env   /opt/conda/envs/turnup_env
 COPY --from=env-unikp     /opt/conda/envs/unikp        /opt/conda/envs/unikp
 COPY --from=env-catapro   /opt/conda/envs/catapro_env  /opt/conda/envs/catapro_env
+COPY --from=env-catpred   /opt/conda/envs/catpred_env  /opt/conda/envs/catpred_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
@@ -174,6 +184,7 @@ RUN find /opt/conda -name "*.pyc" -delete \
 COPY . .
 
 RUN mkdir -p /app/models/EITLEM/Weights \
+             /app/models/CatPred \
              /app/models/TurNup/data/saved_models \
              /app/models/UniKP-main/models \
              /app/media/sequence_info \

api/embeddings/registry.py

@@ -34,11 +34,13 @@
         "implemented": True,
         "conda_env": "esm",
         "python_path_key": "esm2",   # key in config PYTHON_PATHS
-        "used_by": ["KinForm-H", "KinForm-L"],
+        "used_by": ["KinForm-H", "KinForm-L", "CatPred"],
         "notes": (
             "Invoked as a subprocess by KinForm.  The python path is passed via "
             "the KINFORM_ESM_PATH environment variable.  Multi-layer embeddings "
-            "are extracted in a single model-load pass."
+            "are extracted in a single model-load pass. CatPred uses per-residue "
+            "ESM2 features, so it bridges into a method-specific cache format "
+            "rather than reusing the shared mean-vector cache directly."
         ),
     },
 

api/methods/catpred.py

@@ -0,0 +1,50 @@
+# api/methods/catpred.py
+#
+# Method descriptor for CatPred.
+
+from api.methods.base import MethodDescriptor, SubprocessEngineConfig
+
+
+descriptor = MethodDescriptor(
+    key="CatPred",
+    display_name="CatPred",
+    authors=(
+        "Veda Sheersh Boorla, Somtirtha Santra, Costas D. Maranas"
+    ),
+    publication_title=(
+        "CatPred: A comprehensive framework for deep learning in vitro enzyme kinetic parameters"
+    ),
+    citation_url="https://www.nature.com/articles/s41467-025-57215-9",
+    repo_url="https://github.com/maranasgroup/CatPred",
+    more_info=(
+        "CatPred currently integrates kcat and Km through a local CPU inference "
+        "adapter. Ki remains outside the current webKinPred target model."
+    ),
+
+    supports=["kcat", "Km"],
+    input_format="single",
+    output_cols={
+        "kcat": "kcat (1/s)",
+        "Km": "KM (mM)",
+    },
+    max_seq_len=2048,
+    col_to_kwarg={"Substrate": "substrates"},
+    target_kwargs={
+        "kcat": {"kinetics_type": "KCAT"},
+        "Km": {"kinetics_type": "KM"},
+    },
+    subprocess=SubprocessEngineConfig(
+        python_path_key="CatPred",
+        script_key="CatPred",
+        data_path_env={
+            "CATPRED_REPO_ROOT": "CatPred",
+            "CATPRED_MEDIA_PATH": "media",
+            "CATPRED_TOOLS_PATH": "tools",
+            "PYTHONPATH": "CatPred",
+        },
+        extra_env={
+            "PROTEIN_EMBED_USE_CPU": "1",
+        },
+    ),
+    embeddings_used=["esm2"],
+)

docker-compose.prod.yml

@@ -95,6 +95,7 @@ services:
       - ./models/DLKcat/DeeplearningApproach:/app/models/DLKcat/DeeplearningApproach:ro
       - ./models/KinForm/results/trained_models:/app/models/KinForm/results/trained_models:ro
       - ./models/CataPro:/app/models/CataPro:ro
+      - ./models/CatPred:/app/models/CatPred:ro
     environment:
       <<: *common-env
     depends_on:

docker-requirements/catpred_requirements.txt

@@ -0,0 +1,17 @@
+# Requirements for CatPred (Python 3.10)
+torch==2.4.1
+fair-esm==2.0.0
+transformers==4.47.1
+sentencepiece==0.2.0
+numpy==1.26.4
+pandas==2.2.3
+pandas-flavor==0.6.0
+scikit-learn==1.5.2
+scipy==1.14.1
+tqdm==4.67.1
+typed-argument-parser==1.10.1
+descriptastorus==2.8.0
+rotary-embedding-torch==0.6.5
+hyperopt==0.2.7
+matplotlib==3.9.2
+tensorboardX==2.6.2.2

models/CatPred/README.md

@@ -0,0 +1,20 @@
+# CatPred
+
+This directory is reserved for the CatPred integration.
+
+Expected contents:
+- a checkout of the CatPred repository rooted at `models/CatPred/`
+- the adapter entrypoint at `models/CatPred/catpred/integration/webkinpred_adapter.py`
+
+Local development can instead point webKinPred at an external CatPred checkout:
+
+```bash
+export WEBKINPRED_CATPRED_ROOT="/absolute/path/to/CatPred"
+export WEBKINPRED_CATPRED_PYTHON="/absolute/path/to/CatPred/.venv/bin/python"
+```
+
+Docker/runtime notes:
+- `webKinPred/config_docker.py` defaults to `/app/models/CatPred`
+- the CatPred subprocess descriptor sets `CATPRED_REPO_ROOT`, `CATPRED_MEDIA_PATH`, and `CATPRED_TOOLS_PATH`
+- CatPred kcat/Km use per-residue ESM2 features and cache them under
+  `media/sequence_info/esm2_last/per_residue/{seq_id}.pt`

webKinPred/config_base.py

@@ -10,6 +10,7 @@
 
 
 _DATA_PATH_REL = {
+    "CatPred": "models/CatPred",
     "DLKcat": "models/DLKcat/DeeplearningApproach/Data",
     "DLKcat_Results": "models/DLKcat/DeeplearningApproach/Results",
     "EITLEM": "models/EITLEM",
@@ -23,6 +24,7 @@
 
 
 _PREDICTION_SCRIPT_REL = {
+    "CatPred": "models/CatPred/catpred/integration/webkinpred_adapter.py",
     "DLKcat": "models/DLKcat/DeeplearningApproach/Code/example/prediction_for_input.py",
     "EITLEM": "models/EITLEM/Code/eitlem_prediction_script_batch.py",
     "TurNup": "models/TurNup/code/kcat_prediction_batch.py",

webKinPred/config_docker.py

@@ -1,3 +1,5 @@
+import os
+
 from webKinPred.config_base import (
     DEFAULT_ALLOWED_FRONTEND_IPS,
     SERVER_LIMIT,
@@ -13,6 +15,7 @@
 FASTAS_DIR = f"{BASE_PATH}/fastas"
 
 PYTHON_PATHS = {
+    "CatPred": "/opt/conda/envs/catpred_env/bin/python",
     "DLKcat": "/opt/conda/envs/dlkcat_env/bin/python",
     "EITLEM": "/opt/conda/envs/eitlem_env/bin/python",
     "TurNup": "/opt/conda/envs/turnup_env/bin/python",
@@ -28,6 +31,11 @@
 DATA_PATHS = build_data_paths(BASE_PATH)
 PREDICTION_SCRIPTS = build_prediction_scripts(BASE_PATH)
 
+CATPRED_ROOT = os.environ.get("WEBKINPRED_CATPRED_ROOT")
+if CATPRED_ROOT:
+    DATA_PATHS["CatPred"] = CATPRED_ROOT
+    PREDICTION_SCRIPTS["CatPred"] = f"{CATPRED_ROOT}/catpred/integration/webkinpred_adapter.py"
+
 SIMILARITY_DATASETS = build_similarity_datasets(FASTAS_DIR)
 TARGET_DBS = {label: item["target_db"] for label, item in SIMILARITY_DATASETS.items()}
 

webKinPred/config_local.py

@@ -26,6 +26,7 @@ def _env_python(env_name: str) -> str:
 
 
 PYTHON_PATHS = {
+    "CatPred": _env_python("catpred_env"),
     "DLKcat": _env_python("dlkcat_env"),
     "EITLEM": _env_python("eitlem_env"),
     "TurNup": _env_python("turnup_env"),
@@ -41,6 +42,16 @@ def _env_python(env_name: str) -> str:
 DATA_PATHS = build_data_paths(BASE_PATH)
 PREDICTION_SCRIPTS = build_prediction_scripts(BASE_PATH)
 
+CATPRED_ROOT = os.environ.get("WEBKINPRED_CATPRED_ROOT")
+if CATPRED_ROOT:
+    catpred_root = str(Path(CATPRED_ROOT).resolve())
+    DATA_PATHS["CatPred"] = catpred_root
+    PREDICTION_SCRIPTS["CatPred"] = str(Path(catpred_root) / "catpred" / "integration" / "webkinpred_adapter.py")
+    PYTHON_PATHS["CatPred"] = os.environ.get(
+        "WEBKINPRED_CATPRED_PYTHON",
+        _env_python("catpred_env"),
+    )
+
 SIMILARITY_DATASETS = build_similarity_datasets(FASTAS_DIR)
 TARGET_DBS = {label: item["target_db"] for label, item in SIMILARITY_DATASETS.items()}