WESworkflow

A Practical Workflow for Correcting Kit-specific Effects in Whole-Exome Sequencing Data

This repository provides an end-to-end workflow for processing whole-exome sequencing (WES) data, from raw reads to gene-level feature matrices (CADD-weighted Allele Fraction), batch effect assessment, and imputation-based correction.

---

Requirements

• Bash, R (≥4.0), Python (for PARC)

• Tools: BWA, CrossMap (optional), samtools, Picard, DeepVariant, GLnexus, bcftools, Beagle, ANNOVAR

• Additional R packages listed in the R scripts.

• External resources:

  • Reference genome (GRCh38)
  • Reference panel (provided via Zenodo)
  • Genetic maps (for Beagle)
  • ANNOVAR databases
  • CADD prescored database

---

Repository structure

Data/                      # example inputs and auxiliary resources
Data_pre_processing/       # QC + trimming + alignment / liftover
Variant_calling/           # DeepVariant + GLnexus
Variant_post_processing/   # genotype imputation with BEAGLE + ANNOVAR/CADD annotation
Variant_to_gene/           # gene-level variant aggregation
Gene-level imputation/     # clustering, GMM for threshold determination, gene-level imputation imputation

---

Workflow

1. Data preprocessing

Scripts:

Data_pre_processing/QC/trimm.sh
Data_pre_processing/QC/fastqc/
Data_pre_processing/Alignment/alignment.sh
Data_pre_processing/Liftover/liftover_bams.sh

Steps:

• QC and trimming
• alignment to GRCh38
• duplicate marking
• optional liftover

---

2. Variant calling

Scripts:

Variant_calling/Calling/run_deepvariant.sh
Variant_calling/Joint_genotyping/run_GLnexus.sh

Steps:

• per-sample variant calling (DeepVariant)
• joint genotyping (GLnexus)

---

3. Genotype imputation

Script:

Variant_post_processing/1_genotype_imputation.sh

Steps:

• normalization and filtering
• genotype conformation
• imputation (Beagle)

---

4. Variant annotation

Script:

Variant_post_processing/2_annotation.sh

Steps:

• functional annotation (ANNOVAR)
• filtering to coding/splicing variants
• adding CADD scores

---

5. Gene-level feature generation

Scripts:

Variant_to_gene/gene_aggregation.sh
Variant_to_gene/cal_features_multi.py

Steps:

• aggregation of variants per gene
• computation of gene-level features

---

6. Feature analysis and imputation

Scripts:

Gene-level imputation/
├── 1_features_loading.R
├── 2_clustering.R
├── 3_GMM.R
├── 4_feature_imputation.R

Steps:

• feature loading
• clustering (PARC)
• detection rate modeling (GMM) for feature imputation thresholds
• MNAR-aware gene-level KNN imputation

---

Data

• Example metadata:

Data/example/sample_path_map_example.tsv

• BED regions:

Data/bed/refGene_exons_splice5.nochr.bed

• LoF annotations:

Data/gnomad_lofs/

• Reference panel (EUR subset):

  • provided separately via Zenodo (see DOI)

---

Notes

• All scripts use /path/to/... placeholders — paths must be adapted
• Large datasets (VCF, BAM, reference panels) are not included
• Workflow is modular — individual steps can be run independently