Webapp

.github/workflows/deploy-space.yml

@@ -0,0 +1,33 @@
+name: Deploy to HF Spaces
+
+on:
+  push:
+    branches: [main]
+    paths:
+      - "9.webapp/**"
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          lfs: true
+
+      - name: Install huggingface_hub
+        run: pip install huggingface_hub
+
+      - name: Push to HF Spaces
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        run: |
+          cd 9.webapp
+          hf upload WayScience/gene-dependency-explorer . . \
+            --repo-type space \
+            --exclude ".venv/**" \
+            --exclude "**/__pycache__/**" \
+            --exclude "**/*.pyc" \
+            --exclude "**/*.egg-info/**" \
+            --exclude "data/CRISPRGeneEffect.parquet" \
+            --exclude "data/CRISPR_gene_dictionary.parquet" \
+            --commit-message "deploy from ${{ github.sha }}"

.gitignore

@@ -4,3 +4,7 @@
 __pycache__
 2.train-VAE/parameter_sweep
 .DS_Store
+9.webapp/gene_dependency_webapp.egg-info
+9.webapp/data/CRISPRGeneEffect.parquet
+9.webapp/data/CRISPR_gene_dictionary.parquet
+*.pyc

.pre-commit-config.yaml

@@ -0,0 +1,64 @@
+repos:
+-   repo: https://gitlab.com/vojko.pribudic.foss/pre-commit-update
+    rev: v0.9.0 # Insert the latest tag here
+
+    hooks:
+    -   id: pre-commit-update
+        args: [--exclude, black, --keep, isort]
+    # Formats import order
+-   repo: https://github.com/pycqa/isort
+    rev: 5.12.0
+    hooks:
+    -   id: isort
+        name: isort (python)
+        args: ["--profile", "black", "--filter-files"]
+
+    #Code formatter for both python files and jupyter notebooks
+-   repo: https://github.com/astral-sh/ruff-pre-commit
+    # Ruff version.
+    rev: v0.15.17
+    hooks:
+      # Run the formatter.
+    -   id: ruff-format
+
+-   repo: https://github.com/nbQA-dev/nbQA
+    rev: 1.9.1
+    hooks:
+    -   id: nbqa-isort
+        additional_dependencies: [isort==5.6.4]
+        args: [--profile=black]
+
+-   repo: https://github.com/pre-commit/mirrors-prettier
+    rev: "v3.1.0"
+    hooks:
+    -   id: prettier
+        types_or: [markdown, yaml, json, html, css, python, r]
+
+-   repo: https://github.com/hadialqattan/pycln
+    rev: v2.6.0
+    hooks:
+    -   id: pycln
+        args: [--config=pyproject.toml]
+        exclude: |
+          (?x)^(
+            2\.segment_images/scripts/.*\.py$|
+            3\.cellprofiling/scripts/.*\.py$
+          )$
+    # additional hooks found with in the pre-commit lib
+-   repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+    -   id: trailing-whitespace # removes trailing white spaces
+    -   id: mixed-line-ending # removes mixed end of line
+        args:
+        -   --fix=lf
+
+-   repo: local
+    hooks:
+    -   id: shellcheck
+        name: shellcheck
+        description: Test shell scripts with shellcheck
+        entry: shellcheck
+        language: python
+        types: [shell]
+        require_serial: true # shellcheck can detect sourcing this way

5.drug-dependency/9.precompute_webapp_pca_embeddings.ipynb

@@ -0,0 +1,238 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "notice-cell",
+   "metadata": {},
+   "source": [
+    "# Precompute PCA Embeddings for Webapp\n",
+    "\n",
+    "> **Note:** This notebook was a one-time preprocessing step run to generate the pre-computed PCA\n",
+    "> parquet files bundled with the `9.webapp/` Streamlit app. It is **not intended to be re-run**\n",
+    "> without modification.\n",
+    ">\n",
+    "> The helper functions below (`load_model_data`, `single_load_data`, `latent_load_data`) were\n",
+    "> originally defined in `9.webapp/app_utils.py` but were removed from the app once the pre-computed\n",
+    "> parquets replaced runtime PCA computation. They are reproduced here for provenance.\n",
+    ">\n",
+    "> **Dependencies that are no longer in the webapp environment:**\n",
+    "> - `CRISPRGeneEffect.parquet` and `CRISPR_gene_dictionary.parquet` (large gitignored DepMap files\n",
+    ">   — download via `0.data-download/` first)\n",
+    "> - `sklearn` (`scikit-learn`) — not in `9.webapp/pyproject.toml`\n",
+    ">\n",
+    "> **Outputs** (written to `9.webapp/data/`):\n",
+    "> - `pca_embeddings_single_dependencies.parquet`\n",
+    "> - `pca_embeddings_latent_reactome.parquet`\n",
+    "> - `pca_embeddings_latent_corum.parquet`\n",
+    "> - `pca_embeddings_latent_drug.parquet`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "6c08a8f0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pathlib\n",
+    "\n",
+    "import pandas as pd\n",
+    "from sklearn.decomposition import PCA"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "paths-cell",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Paths — run from repo root or adjust accordingly\n",
+    "REPO_ROOT = pathlib.Path().resolve().parent  # goes up from 5.drug-dependency/ to repo root\n",
+    "WEBAPP_DIR = REPO_ROOT / \"9.webapp\"\n",
+    "DATA_DIR = WEBAPP_DIR / \"data\"\n",
+    "RESULTS_DIR = REPO_ROOT / \"5.drug-dependency\" / \"results\"\n",
+    "\n",
+    "print(f\"Repo root:   {REPO_ROOT}\")\n",
+    "print(f\"Webapp data: {DATA_DIR}\")\n",
+    "print(f\"Results dir: {RESULTS_DIR}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "helpers-header",
+   "metadata": {},
+   "source": [
+    "## Helper functions\n",
+    "\n",
+    "Originally from `9.webapp/app_utils.py`, reproduced here for provenance."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "load-model-data-cell",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def load_model_data(dependency_file, gene_dict_file):\n",
+    "    \"\"\"Load and preprocess gene dependency data and gene dictionary.\"\"\"\n",
+    "    dependency_df = pd.read_parquet(dependency_file)\n",
+    "    print(dependency_df.shape)\n",
+    "\n",
+    "    gene_dict_df = (\n",
+    "        pd.read_parquet(gene_dict_file).query(\"qc_pass\").reset_index(drop=True)\n",
+    "    )\n",
+    "    gene_dict_df.entrez_id = gene_dict_df.entrez_id.astype(str)\n",
+    "\n",
+    "    entrez_genes = [\n",
+    "        x[1].strip(\")\").strip()\n",
+    "        for x in dependency_df.iloc[:, 1:].columns.str.split(\"(\")\n",
+    "    ]\n",
+    "    entrez_intersection = list(\n",
+    "        set(gene_dict_df.entrez_id).intersection(set(entrez_genes))\n",
+    "    )\n",
+    "    gene_dict_df = gene_dict_df.set_index(\"entrez_id\").reindex(entrez_intersection)\n",
+    "\n",
+    "    dependency_df.columns = [\"ModelID\"] + entrez_genes\n",
+    "    dependency_df = dependency_df.loc[:, [\"ModelID\"] + gene_dict_df.index.tolist()]\n",
+    "    dependency_df.columns = [\"ModelID\"] + gene_dict_df.symbol_id.tolist()\n",
+    "    dependency_df = dependency_df.dropna(axis=\"columns\")\n",
+    "    return dependency_df, gene_dict_df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "single-load-cell",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def single_load_data():\n",
+    "    \"\"\"Load CRISPR gene dependency scores merged with cell line metadata.\"\"\"\n",
+    "    dependency_file = DATA_DIR / \"CRISPRGeneEffect.parquet\"\n",
+    "    gene_dict_file = DATA_DIR / \"CRISPR_gene_dictionary.parquet\"\n",
+    "    cancer_type_input_file = DATA_DIR / \"Model.parquet\"\n",
+    "\n",
+    "    dependency_df, _ = load_model_data(dependency_file, gene_dict_file)\n",
+    "    dependency_df = dependency_df.set_index(\"ModelID\")\n",
+    "    cancer_type_df = pd.read_parquet(cancer_type_input_file)\n",
+    "    combined_df = dependency_df.merge(\n",
+    "        cancer_type_df[[\"ModelID\", \"OncotreePrimaryDisease\"]], on=\"ModelID\", how=\"left\"\n",
+    "    )\n",
+    "    return combined_df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "latent-load-cell",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def latent_load_data():\n",
+    "    \"\"\"Load latent score parquets and pivot to wide format for PCA.\"\"\"\n",
+    "    reactome_df = pd.read_parquet(RESULTS_DIR / \"all_reactome_results.parquet\")\n",
+    "    corum_df = pd.read_parquet(RESULTS_DIR / \"all_corum_results.parquet\")\n",
+    "    drug_df = pd.read_parquet(RESULTS_DIR / \"all_drug_results.parquet\")\n",
+    "\n",
+    "    reactome_df[\"feature\"] = reactome_df[\"reactome_pathway\"]\n",
+    "    reactome_df = reactome_df.drop(\n",
+    "        columns=[\"model\", \"latent_dim_total\", \"init\", \"seed\", \"z\", \"reactome_pathway\"]\n",
+    "    )\n",
+    "    corum_df[\"feature\"] = corum_df[\"reactome_pathway\"]\n",
+    "    corum_df = corum_df.drop(\n",
+    "        columns=[\"model\", \"latent_dim_total\", \"init\", \"seed\", \"z\", \"reactome_pathway\"]\n",
+    "    )\n",
+    "    drug_df[\"feature\"] = drug_df[\"name\"]\n",
+    "    drug_df = drug_df.drop(\n",
+    "        columns=[\"model\", \"latent_dim_total\", \"init\", \"seed\", \"z\", \"name\"]\n",
+    "    )\n",
+    "\n",
+    "    meta_cols = [\"ModelID\", \"OncotreePrimaryDisease\"]\n",
+    "    reactome_meta = reactome_df[meta_cols].drop_duplicates()\n",
+    "    corum_meta = corum_df[meta_cols].drop_duplicates()\n",
+    "    drug_meta = drug_df[meta_cols].drop_duplicates()\n",
+    "\n",
+    "    reactome_matrix = reactome_df.pivot(\n",
+    "        index=\"ModelID\", columns=\"feature\", values=\"latent_score\"\n",
+    "    ).merge(reactome_meta, on=\"ModelID\", how=\"left\")\n",
+    "    corum_matrix = corum_df.pivot(\n",
+    "        index=\"ModelID\", columns=\"feature\", values=\"latent_score\"\n",
+    "    ).merge(corum_meta, on=\"ModelID\", how=\"left\")\n",
+    "    drug_matrix = drug_df.pivot(\n",
+    "        index=\"ModelID\", columns=\"feature\", values=\"latent_score\"\n",
+    "    ).merge(drug_meta, on=\"ModelID\", how=\"left\")\n",
+    "\n",
+    "    return reactome_matrix, corum_matrix, drug_matrix"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "pca-header",
+   "metadata": {},
+   "source": [
+    "## Compute and save PCA embeddings"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "6cdc8f80",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def compute_pca(df, metadata_columns) -> pd.DataFrame:\n",
+    "    print(df.shape[0], \"rows before dropping NA\")\n",
+    "    df = df.dropna()\n",
+    "    print(df.shape[0], \"rows after dropping NA\")\n",
+    "    metadata_df = df[metadata_columns].copy()\n",
+    "    feat_cols = df.columns.drop(metadata_columns)\n",
+    "    if \"source\" in feat_cols:\n",
+    "        feat_cols = feat_cols.drop(\"source\")\n",
+    "    pca_input = df[feat_cols].apply(pd.to_numeric, errors=\"coerce\")\n",
+    "    pca = PCA(n_components=2, random_state=0)\n",
+    "    pca_embedding = pca.fit_transform(pca_input)\n",
+    "    metadata_df[\"PCA1\"] = pca_embedding[:, 0]\n",
+    "    metadata_df[\"PCA2\"] = pca_embedding[:, 1]\n",
+    "    return metadata_df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "run-pca-cell",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "combined_df = single_load_data()\n",
+    "reactome_matrix, corum_matrix, drug_matrix = latent_load_data()\n",
+    "\n",
+    "pca_single_df = compute_pca(combined_df, [\"ModelID\", \"OncotreePrimaryDisease\"])\n",
+    "pca_latent_reactome_df = compute_pca(reactome_matrix, [\"OncotreePrimaryDisease\", \"ModelID\"])\n",
+    "pca_latent_corum_df = compute_pca(corum_matrix, [\"OncotreePrimaryDisease\", \"ModelID\"])\n",
+    "pca_latent_drug_df = compute_pca(drug_matrix, [\"OncotreePrimaryDisease\", \"ModelID\"])\n",
+    "\n",
+    "pca_single_df.to_parquet(DATA_DIR / \"pca_embeddings_single_dependencies.parquet\")\n",
+    "pca_latent_reactome_df.to_parquet(DATA_DIR / \"pca_embeddings_latent_reactome.parquet\")\n",
+    "pca_latent_corum_df.to_parquet(DATA_DIR / \"pca_embeddings_latent_corum.parquet\")\n",
+    "pca_latent_drug_df.to_parquet(DATA_DIR / \"pca_embeddings_latent_drug.parquet\")\n",
+    "\n",
+    "print(\"Done — parquets written to\", DATA_DIR)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "name": "python",
+   "version": "3.12.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

9.webapp/.hfignore

@@ -0,0 +1,5 @@
+.venv/
+__pycache__/
+*.pyc
+*.egg-info/
+.DS_Store

9.webapp/.streamlit/config.toml

@@ -0,0 +1,2 @@
+[server]
+fileWatcherType = "none"

9.webapp/Dockerfile

@@ -0,0 +1,18 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+COPY --from=ghcr.io/astral-sh/uv:latest /uv /usr/local/bin/uv
+
+COPY pyproject.toml uv.lock ./
+RUN uv sync --frozen --no-dev
+
+COPY . .
+
+EXPOSE 7860
+
+CMD ["uv", "run", "streamlit", "run", "app.py", \
+     "--server.address", "0.0.0.0", \
+     "--server.port", "7860", \
+     "--server.headless", "true", \
+     "--server.fileWatcherType", "none"]