VCF → annotation → perturbation → pathways → phenotype

1. download VCF
2. annotate variants (VEP + SnpEff)
3. measure perturbations (CPA trained/fine-tuned on X-atlas, factoring in personal variants)
4. model disregulated pathways via chain reactions (PathDNN + Reactome)
5. measure phenotype differences (DeepDR + GWAS Catalog)
6. return visualized results (make it beautiful)

1. upload VCF file (Nucleus, Nebula Genomics)
2. user sees progress bars for:
   • variant annotation (VEP/SnpEff)
   • functional simulation (CPA)
   • pathway mapping (PathDNN)
   • phenotype inference (DeepDR)
   • report generation
3. a report shows how their unique variants ripple through genes, pathways, and traits

1. VCF is uploaded
2. annotate variants (VEP + SnpEff)
3. measure perturbations (CPA trained/fine-tuned on X-atlas, factoring in personal variants)
4. model disregulated pathways via chain reactions (PathDNN + Reactome)
5. measure phenotype differences (DeepDR + GWAS Catalog)
6. return results to user in pretty viualization

Packages:

• Python 3.10
• PyTorch 2.6.0 with CUDA 12.4
• scanpy & anndata (single-cell genomics analysis)
• pandas & numpy
• scikit-learn
• cyvcf2 (VCF file processing)
• networkx (graph analysis)
• matplotlib & seaborn
• scvi-tools (single-cell variational inference)

# 1. activate the environment
conda activate murmur

# 2. fill out machine config
python ./config/generate_config.sh

# 3. run pipeline
python -m src.pipeline

expr_pred, pathway_scores, trait_scores = run(
    vcf_path="path/to/sample.vcf",
    out_dir="results/",
    xatlas_h5ad="data/xatlas.h5ad",
    msigdb_gmt="data/msigdb.gmt",
    gwas_map="data/gwas_traits.tsv"
)

# VEP annotation
# allow the docker container to write to local /results folder (not most secure practice)
chmod 777 results/

# run VEP on your VCF file
sudo docker run --rm \
  -v {{VCF_DIR}}:/data/input \
  -v {{PROJECT_DIR}}/results:/data/output \
  ensemblorg/ensembl-vep:release_109.3 \
  vep --input_file /data/input/{{VCF_FILENAME}} \
  --output_file /data/output/{{OUTPUT_FILENAME}} \
  --database --assembly=GRCh38 --tab --no_stats --force_overwrite

• {{VCF_DIR}}: directory containing your VCF file
• {{VCF_FILENAME}}: name of your VCF file (ie sample.vcf.gz)
• {{OUTPUT_FILENAME}}: desired output name (ie annotated_variants.tsv)
• {{PROJECT_DIR}}: path to this murmur project directory