A complete, interactive platform for performing QC, GWAS, ML modeling, gene mapping, enrichment, and predictive analysis — powered by Shiny and ready for HPC deployment via Singularity.

1. Features
2. Screenshots
3. Tech Stack
4. Installation
   • Run Locally
   • Run with Singularity (HPC)
5. How to Use
6. Downloadable Outputs
7. Common Errors
8. Credits & Contact

• 📤 Upload your own VCF, phenotype, and covariate data
• ✅ Perform SNP Quality Control with MAF, HWE, and Call Rate thresholds
• 📊 Explore data via PCA and UMAP projections
• 🔬 Run GWAS with Bonferroni option and adjustable p-value slider
• 🧪 View Q-Q and Manhattan plots with chromosome/position filters
• 🧬 Map SNPs to nearest human genes
• 🧭 Perform Pathway Enrichment Analysis with barplots for KEGG, GO, and Reactome
• 🤖 Train/test a Random Forest model using significant SNPs
• 📈 Inspect model predictions, AUC, sensitivity/specificity, and variable importance
• 📉 Run power analysis with effect size and power curves
• 🧾 Export all results as CSV

• R 4.x via rocker/shiny
• Shiny + data.table, DT, plotly, ranger, mlr3, VariantAnnotation, enrichR, pwr, ROCR
• Containerized using Singularity
• Designed for HPC cluster-friendly use

1. Clone this repo: in `bash git clone https://github.com/jcaperella29/GWAS_SHINY_APP cd GWAS_SHINY_APP

2.Launch the app in R

shiny::runApp()

Make sure required R packages are installed (listed in app.R)

###⚗️ Run with Singularity and SLURM (HPC)

1.Build container:

in bash sudo singularity build gwas_app.sif Singularity.def

2. Submit the app to SLURM Use the included SLURM batch script to launch the Shiny app in a containerized HPC environment:

in bash sbatch run_gwas_app.slurm

This will:

Launch the app inside the container

Bind it to 0.0.0.0 on port 3838

Generate a log file: gwas_shiny.log

3. Access the app If you're on WSL or a local machine:

Open your browser and go to: http://localhost:3838 If you're running on a remote VM:

ssh -L 3838:localhost:3838 your_user@your_server_ip

Then visit:

http://localhost:3838

🧭 How to Use Upload Data Upload VCF, phenotype (CSV), and optional covariates (CSV)

Run QC Click Run QC button after adjusting MAF/HWE/Call Rate thresholds

Explore Data Visualize sample space with PCA and UMAP

Power Analysis Hit Run Power to estimate statistical power and see the power curve

Run GWAS Adjust the p-value slider and choose Bonferroni if needed, then click Run GWAS

Visualize GWAS Use Q-Q plot and Manhattan plot (adjust chr/position filters as needed)

Map SNPs to Genes Click Map Genes and view table + plot

Run Enrichment Perform pathway enrichment and view top results via Barplot

Run Random Forest Train/test classifier on final hits. View predictions, metrics, and SNP importance

Download Outputs Use sidebar download buttons to get CSVs of all results

📥 Downloadable Outputs GWAS full results

Covariate associations

Final phenotype-only hits

SNP-to-gene annotation

Pathway enrichment table

Power analysis summary

Random forest predictions, metrics, and feature importance

❌ Common Errors Error Cause Solution Not enough samples in one or both groups Trait imbalance Ensure both classes are represented Can't read output 'rfPredictions' RF model not trained Click Run Random Forest first Gene mapping failed Invalid SNP coordinates or gene DB issue Check VCF chr formatting (should match hg19) Enrichment failed Too few gene symbols Ensure gene mapping produced valid symbols Package not found Singularity not built properly Rebuild container with correct R packages

🙋‍♂️ Credits & Contact 🔬 App by John Caperella

📬 Contact: @jcaperella29 “Stealing the distorted desires of legacy pipelines.” 🎩

Ready for GWAS evolution.