fix: DIG stage writes bmr_pmfs.dig.csv directly to the cohort dir

The DIG provider names its output bmr_pmfs.dig.csv (not bmr_pmfs.csv) and needs the
out dir to exist; point it at output/pancan/<C> directly (no temp dir / cp), which does
not clobber CBaSE's bmr_pmfs.csv. Validated on ACC: CBaSE + MutSig2CV (Docker) both run
cleanly on the cBioPortal pancan MAF; DIG produces a 21MB per-gene PMF.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: ahmed-shuaibi

scripts/run_cohort_pipeline.sh

@@ -32,14 +32,11 @@ else log "skip cbase"; fi
 N=0
 [ -f "${ROOT}/${C}/count_matrix.csv" ] && N=$(( $(wc -l < "${ROOT}/${C}/count_matrix.csv") - 1 ))
 
-# 2. DIG --------------------------------------------------------------------
+# 2. DIG (writes bmr_pmfs.dig.csv directly; does not clobber CBaSE's bmr_pmfs.csv) ---
 if [ ! -f "${ROOT}/${C}/bmr_pmfs.dig.csv" ] && [ "$N" -gt 0 ]; then
   log "DIG generate (N=${N})"
-  rm -rf "/tmp/dig_${C}"
-  if "$DIALECT" generate -m "$MAF" -o "/tmp/dig_${C}" --bmr dig \
-       --dig-results "$DIG_RESULTS" --dig-samples "$N" -r hg19; then
-    cp "/tmp/dig_${C}/bmr_pmfs.csv" "${ROOT}/${C}/bmr_pmfs.dig.csv"
-  else log "STAGE-FAIL dig"; fi
+  "$DIALECT" generate -m "$MAF" -o "${ROOT}/${C}" --bmr dig \
+    --dig-results "$DIG_RESULTS" --dig-samples "$N" -r hg19 || log "STAGE-FAIL dig"
 else log "skip dig"; fi
 
 # 3. MutSig2CV (Docker) -----------------------------------------------------