Fix critical bugs, improve performance, add Snakemake workflow + CI/CD

.github/workflows/ci.yml

@@ -0,0 +1,95 @@
+name: CI
+
+on:
+  push:
+    branches: [main, full-pipeline]
+  pull_request:
+    branches: [main, full-pipeline]
+  workflow_dispatch:
+
+jobs:
+  # ---------------------------------------------------------------------
+  # Fast, no-network-deps check: every .py file parses with base Python
+  # (catches typos/syntax errors immediately, no package installs needed).
+  # ---------------------------------------------------------------------
+  syntax-check:
+    name: Python syntax check
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - name: Parse every .py file
+        run: |
+          set -e
+          status=0
+          while IFS= read -r -d '' f; do
+            python -m py_compile "$f" || status=1
+          done < <(find . -name "*.py" -not -path "./.git/*" -print0)
+          exit $status
+
+  # ---------------------------------------------------------------------
+  # Validate the Snakemake DAG for every variant/config combination,
+  # without needing RNAfold, the C++ binary, or trained models.
+  # ---------------------------------------------------------------------
+  snakemake-dry-run:
+    name: Snakemake DAG validation
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - name: Install Snakemake
+        run: |
+          python -m pip install --upgrade pip
+          pip install snakemake "pulp<2.8"
+      - name: Dry-run every variant / config combination
+        run: |
+          set -e
+          snakemake --cores 1 -n
+          snakemake --cores 1 -n --config run_scoring=true
+          snakemake --cores 1 -n --config variant=broad_pam
+          snakemake --cores 1 -n --config variant=base_editor
+      - name: Reject an invalid variant
+        run: '! snakemake --cores 1 -n --config variant=bogus'
+
+  # ---------------------------------------------------------------------
+  # Real execution smoke test: installs the actual Python dependencies
+  # (numpy/pandas/biopython/ViennaRNA Python bindings -- NOT TensorFlow,
+  # NOT the C++ binary) and runs Steps 1-3 against the bundled demo FASTA.
+  # This is as far as this pipeline can run without the missing
+  # connection_to_matrix.cpp source and trained .h5 models (see README
+  # "Known Issues") -- those two gaps are the repository owner's to fill.
+  # ---------------------------------------------------------------------
+  smoke-test:
+    name: Steps 1-3 smoke test (demo FASTA)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - name: Install lightweight pipeline dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install numpy pandas biopython ViennaRNA
+          pip install snakemake "pulp<2.8"
+      - name: Run steps 1-3 against the demo FASTA
+        run: snakemake --cores 1 Structure_file.csv Structure_Connection.fa Target_sequence_feature.csv
+      - name: Verify outputs
+        run: |
+          for f in Structure_file.csv Structure_Connection.fa Structure_Connection.csv Target_sequence_feature.csv; do
+            test -s "$f" || { echo "Missing or empty output: $f"; exit 1; }
+          done
+      - name: Upload smoke-test outputs
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: dgd-smoke-test-outputs
+          path: |
+            *.csv
+            *.fa
+            logs/
+          retention-days: 7

.gitignore

@@ -0,0 +1,42 @@
+# Byte-compiled / cache
+__pycache__/
+*.py[cod]
+*$py.class
+
+# Virtual environments
+venv/
+.venv/
+env/
+
+# Compiled C++ binary (built locally via `make`)
+connection_to_matrix
+
+# Trained model weights (large binaries; not tracked in git)
+models/
+*.h5
+
+# Generated pipeline intermediate/output files (see PipelineFiles in DGD.py)
+Structure_file.csv
+Structure_Connection.fa
+Structure_Connection.csv
+Structure_Connection.out
+Structure_Connection.outs
+Structure_Cas9_out.txt
+Structure_out.txt
+Structure_basepairs.csv
+spacer_scaffold_basepairs.csv
+spacer_scaffold_feature.csv
+Structural_annotation.csv
+Feature_Data_Spacer_Scaffold.csv
+Deep_learning_file.csv
+DGD.csv
+DGDVar.csv
+DGDBE.csv
+results/
+
+# Snakemake
+.snakemake/
+
+# OS / editor cruft
+.DS_Store
+*.swp

Base-Editors/DGDbaseeditor.py

@@ -452,7 +452,18 @@ def _count_gc_content(sequence: str, partner_index: list) -> int:
 
 def _calculate_free_energy(dimers: list) -> float:
     stacking = return_stacking_model()
-    return sum(stacking[dimers[i]][dimers[i+1][::-1]] for i in range(0, len(dimers), 2))
+    # See KNOWN ISSUE note in stacking_model.return_stacking_model(): the
+    # table only covers 4 of 16 dinucleotide contexts, and `dimers` is not
+    # guaranteed to have an even length. Fall back to 0.0 / skip a trailing
+    # unpaired entry instead of crashing (a plain dict lookup here would
+    # raise KeyError/IndexError on real sequences).
+    energy = 0.0
+    for i in range(0, len(dimers) - 1, 2):
+        context = stacking.get(dimers[i])
+        if context is None:
+            continue
+        energy += context.get(dimers[i + 1][::-1], 0.0)
+    return energy
 
 
 def compute_final_features(
@@ -470,7 +481,7 @@ def compute_final_features(
         sequence, structure = vals[0], vals[1]
         p_idx  = adjust_partner_index(return_partner_index(structure))
         _,count= return_connection_length(p_idx)
-        dimers = return_dimers_array(sequence, p_idx)
+        dimers, _anti_seq = return_dimers_array(sequence, p_idx)  # BUG FIX: was assigned the raw (dimers, anti_seq) tuple
         gc     = _count_gc_content(sequence, p_idx) / count if count else 0.0
         ids_gc[seq_id]      = [gc, _calculate_free_energy(dimers)]
         ids_mono_di[seq_id] = [_count_monomer(sequence, p_idx), _count_dimer(dimers)]