How To Analyze Whole Exome Sequencing Data For Absolute Beginners: From Raw Reads to High-Quality Variants, Mutations, and CNVs

November 2, 2025

Lei

Table of Contents

Introduction: Understanding Whole Exome Sequencing vs. Whole Genome Sequencing

What is Whole Exome Sequencing (WES)?

Whole Exome Sequencing (WES) is a targeted sequencing approach that focuses specifically on the protein-coding regions of the genome, known as exons. While the human genome contains approximately 3.2 billion base pairs, the exome represents only about 1-2% of this total (~30-50 million base pairs). However, this small fraction contains roughly 85% of known disease-causing variants, making WES an extremely powerful and cost-effective approach for medical genomics.

Think of it this way: If the genome were a massive encyclopedia, WES would be like reading only the most important chapters that contain the core information, while WGS reads every single page including footnotes, appendices, and blank spaces.

Key Differences Between WGS and WES

Understanding these fundamental differences is crucial for choosing the right approach and interpreting results correctly:

Aspect	Whole Genome Sequencing (WGS)	Whole Exome Sequencing (WES)
Target Coverage	Entire genome (~3.2 billion bp)	Exons only (~30-50 million bp)
Cost per Sample	$800-2,000	$200-500
Data Size	30-100 GB per sample	2-8 GB per sample
Typical Coverage Depth	30-50x	100-200x
Analysis Time	8-24 hours	2-6 hours
Variants Detected	4-5M SNVs, 0.5-1M indels	20-25K coding variants
CNV Detection	Excellent (all sizes)	Limited (exon-level resolution)
Regulatory Variants	All regulatory regions	Misses most regulatory variants
Clinical Diagnostic Yield	35-40%	25-30%

When to Choose WES vs. WGS

Choose WES when:

Budget is a primary concern – WES costs 3-4x less than WGS
Suspected Mendelian disorders – Most are caused by coding variants
Large population studies – Cost savings enable larger sample sizes
Clinical diagnostics focused on known disease genes
Rapid turnaround needed – Smaller datasets process faster
Pharmacogenomics – Most drug-related variants are in coding regions

Choose WGS when:

Comprehensive variant discovery is required
Structural variants and CNVs are important
Regulatory variants may be disease-relevant
Previous WES was negative but genetic cause is still suspected
Research applications requiring complete genomic picture
Complex inheritance patterns need investigation

The WES Workflow: What’s Different?

This tutorial focuses specifically on the differences between WES and WGS analysis. The fundamental steps remain similar, but key modifications include:

Target Region Definition – Using capture kit target files
Quality Control – Emphasis on on-target rate and coverage uniformity
Variant Calling – Restricted to captured regions for efficiency
Somatic Mutation Calling – Optimized for cancer genomics applications
CNV Detection – Specialized approaches for targeted data
Interpretation – Focus on protein-coding consequences

Important: This tutorial assumes you’ve already reviewed our WGS analysis series. We’ll reference those tutorials and focus only on WES-specific modifications.

Prerequisites and Environment Setup

Required Background

Before starting this WES tutorial, ensure you have:

Completed our WGS analysis foundation: Part 1: From Raw Reads to High-Quality Variants Using GATK
Basic understanding of variant calling from the WGS tutorial
Familiarity with GATK tools and command-line operations
The conda environment (wgs_analysis) from Part 1 already set up

WES-Specific Software Installation

# Activate the existing WGS analysis environment from Part 1
conda activate wgs_analysis

WES Project Directory Structure

# Create organized directory structure for WES analysis
mkdir -p ~/wes_analysis/{data,references,results,scripts}
mkdir -p ~/wes_analysis/results/{qc,trimmed,aligned,variants,somatic,cnv,annotation}

# Display the working directory structure
tree ~/wes_analysis -L 2

# Navigate to working directory
cd ~/wes_analysis

Key Difference #1: Target Region Files and Reference Preparation

The most fundamental difference in WES analysis is the use of target region files that define which genomic regions were captured during library preparation.

Understanding Capture Technologies

WES uses biotinylated probes (baits) to selectively capture exonic regions before sequencing. Different capture kits target slightly different sets of exons:

Agilent SureSelect – Uses RNA baits, excellent uniformity
Illumina Nextera – DNA baits, good for degraded samples
Roche SeqCap – DNA baits, customizable designs
Twist Bioscience – Synthetic DNA baits, latest technology

Critical Point: Always use the exact target region file that matches your capture kit. Using the wrong target file will lead to incorrect quality metrics and suboptimal results.

Understanding Target Region Formats

Different tools require different formats. Here’s what each file contains:

BED Format (exome_targets.bed):

chr1    69090   70008   
chr1    367658  368597
chr1    621095  622034

GATK Interval List Format (exome_targets.interval_list):

# Create interval list for GATK (required format for some tools)
gatk BedToIntervalList \
    -I exome_targets.bed \
    -O exome_targets.interval_list \
    -SD ~/wgs_analysis/reference/Homo_sapiens_assembly38.dict

# GATK Interval List Format
# @HD    VN:1.6
# @SQ    SN:chr1 LN:248956422
# chr1:69091-70008
# chr1:367659-368597

Key Difference #2: WES-Specific Quality Control Metrics

WES quality control includes all standard metrics from WGS plus several targeted sequencing-specific metrics.

Standard QC Steps

The initial quality control steps are identical to our WGS tutorial Part 1.

# Set up variables for WES analysis
SAMPLE="exome_sample1"
TARGET_REGIONS="~/wes_analysis/references/exome_targets.bed"
REFERENCE="~/wgs_analysis/reference/Homo_sapiens_assembly38.fasta"
THREADS=16

# FastQC analysis - identical to WGS workflow
fastqc \
    ~/wes_analysis/data/${SAMPLE}_R1.fastq.gz \
    ~/wes_analysis/data/${SAMPLE}_R2.fastq.gz \
    -o ~/wes_analysis/results/qc/

# Adapter trimming - same parameters as WGS
trim_galore \
    --paired \
    --quality 20 \
    --length 50 \
    --fastqc \
    --output_dir ~/wes_analysis/results/trimmed/ \
    ~/wes_analysis/data/${SAMPLE}_R1.fastq.gz \
    ~/wes_analysis/data/${SAMPLE}_R2.fastq.gz

# Alignment - same BWA-MEM algorithm as WGS
bwa mem \
    -t ${THREADS} \
    -M \
    -R "@RG\tID:${SAMPLE}\tSM:${SAMPLE}\tPL:ILLUMINA\tLB:${SAMPLE}_lib" \
    ${REFERENCE} \
    ~/wes_analysis/results/trimmed/${SAMPLE}_R1_val_1.fq.gz \
    ~/wes_analysis/results/trimmed/${SAMPLE}_R2_val_2.fq.gz \
    | samtools sort -@ ${THREADS} -o ~/wes_analysis/results/aligned/${SAMPLE}.bam

# Mark duplicates - identical to WGS workflow
gatk MarkDuplicates \
    -I ~/wes_analysis/results/aligned/${SAMPLE}.bam \
    -O ~/wes_analysis/results/aligned/${SAMPLE}_marked_duplicates.bam \
    -M ~/wes_analysis/results/aligned/${SAMPLE}_duplicate_metrics.txt \
    --CREATE_INDEX true

# Base quality score recalibration - same as WGS
gatk BaseRecalibrator \
    -I ~/wes_analysis/results/aligned/${SAMPLE}_marked_duplicates.bam \
    -R ${REFERENCE} \
    --known-sites ~/wgs_analysis/reference/Homo_sapiens_assembly38.dbsnp138.vcf.gz \
    --known-sites ~/wgs_analysis/reference/Mills_and_1000G_gold_standard.indels.hg38.vcf.gz \
    -O ~/wes_analysis/results/aligned/${SAMPLE}_recal_data.table

gatk ApplyBQSR \
    -I ~/wes_analysis/results/aligned/${SAMPLE}_marked_duplicates.bam \
    -R ${REFERENCE} \
    --bqsr-recal-file ~/wes_analysis/results/aligned/${SAMPLE}_recal_data.table \
    -O ~/wes_analysis/results/aligned/${SAMPLE}_recalibrated.bam

WES-Specific Quality Metrics

Now we add the targeted sequencing-specific quality assessments:

# Calculate on-target rate and coverage statistics
gatk CollectHsMetrics \
    -I ~/wes_analysis/results/aligned/${SAMPLE}_recalibrated.bam \
    -O ~/wes_analysis/results/qc/${SAMPLE}_hs_metrics.txt \
    -R ${REFERENCE} \
    -BAIT_INTERVALS ~/wes_analysis/references/exome_targets.interval_list \
    -TARGET_INTERVALS ~/wes_analysis/references/exome_targets.interval_list

# Calculate detailed coverage statistics per target
gatk DepthOfCoverage \
    -R ${REFERENCE} \
    -I ~/wes_analysis/results/aligned/${SAMPLE}_recalibrated.bam \
    -O ~/wes_analysis/results/qc/${SAMPLE}_coverage \
    -L ${TARGET_REGIONS} \
    --summaryCoverageThreshold 10 \
    --summaryCoverageThreshold 20 \
    --summaryCoverageThreshold 30 \
    --summaryCoverageThreshold 50 \
    --summaryCoverageThreshold 100

Understanding WES Quality Metrics

Critical WES-Specific Metrics:

1. ON_TARGET_BASES: Percentage of sequenced bases that fall within target regions

Excellent: >80%
Good: 70-80%
Poor: <70%

2. MEAN_TARGET_COVERAGE: Average coverage depth across target regions

Clinical applications: >100x
Research applications: >50x

3. PCT_TARGET_BASES_10X/20X/30X: Percentage of targets covered at various depths

Clinical quality: >95% at 20x coverage

4. FOLD_ENRICHMENT: How much more coverage targets receive vs. genome average

Good enrichment: >40-fold
Poor enrichment: <20-fold

5. AT_DROPOUT and GC_DROPOUT: Coverage uniformity across different GC content

Good uniformity: <10% dropout
Problematic: >20% dropout

Key Difference #3: Targeted Variant Calling

While the variant calling algorithms remain the same, WES analysis restricts calling to target regions for efficiency and accuracy.

HaplotypeCaller with Target Focus

# HaplotypeCaller with target region restriction (much faster than WGS)
gatk HaplotypeCaller \
    -R ${REFERENCE} \
    -I ~/wes_analysis/results/aligned/${SAMPLE}_recalibrated.bam \
    -O ~/wes_analysis/results/variants/${SAMPLE}.g.vcf.gz \
    -ERC GVCF \
    -L ${TARGET_REGIONS} \
    --dbsnp ~/wgs_analysis/reference/Homo_sapiens_assembly38.dbsnp138.vcf.gz

# Convert GVCF to VCF for single-sample analysis
gatk GenotypeGVCFs \
    -R ${REFERENCE} \
    -V ~/wes_analysis/results/variants/${SAMPLE}.g.vcf.gz \
    -O ~/wes_analysis/results/variants/${SAMPLE}_raw_variants.vcf.gz \
    -L ${TARGET_REGIONS}

WES-Optimized Variant Filtering

Higher coverage in WES allows for more stringent quality filters:

# Apply variant filters optimized for high-coverage WES data
gatk VariantFiltration \
    -R ${REFERENCE} \
    -V ~/wes_analysis/results/variants/${SAMPLE}_raw_variants.vcf.gz \
    --filter-expression "QD < 2.0" --filter-name "LowQD" \
    --filter-expression "FS > 60.0" --filter-name "HighFS" \
    --filter-expression "MQ < 40.0" --filter-name "LowMQ" \
    --filter-expression "SOR > 3.0" --filter-name "HighSOR" \
    --filter-expression "MQRankSum < -12.5" --filter-name "LowMQRankSum" \
    --filter-expression "ReadPosRankSum < -8.0" --filter-name "LowReadPosRankSum" \
    --filter-expression "DP < 20" --filter-name "LowDepth" \
    --filter-expression "DP > 500" --filter-name "HighDepth" \
    -O ~/wes_analysis/results/variants/${SAMPLE}_filtered.vcf.gz

# Extract high-quality variants
bcftools view -f PASS -Oz \
    -o ~/wes_analysis/results/variants/${SAMPLE}_pass.vcf.gz \
    ~/wes_analysis/results/variants/${SAMPLE}_filtered.vcf.gz

bcftools index ~/wes_analysis/results/variants/${SAMPLE}_pass.vcf.gz

Multi-Sample WES Variant Calling

For cohort analysis, joint calling is even more beneficial in WES due to higher coverage depth:

# Create sample list
ls ~/wes_analysis/results/variants/*.g.vcf.gz > gvcf_list.txt

# Combine GVCFs (faster due to target restriction)
gatk CombineGVCFs \
    -R ${REFERENCE} \
    --variant gvcf_list.txt \
    -L ${TARGET_REGIONS} \
    -O ~/wes_analysis/results/variants/cohort.g.vcf.gz

# Joint genotyping across the cohort
gatk GenotypeGVCFs \
    -R ${REFERENCE} \
    -V ~/wes_analysis/results/variants/cohort.g.vcf.gz \
    -L ${TARGET_REGIONS} \
    -O ~/wes_analysis/results/variants/cohort_raw_variants.vcf.gz

# Apply cohort-level filtering
gatk VariantFiltration \
    -R ${REFERENCE} \
    -V ~/wes_analysis/results/variants/cohort_raw_variants.vcf.gz \
    --filter-expression "QD < 2.0" --filter-name "LowQD" \
    --filter-expression "FS > 60.0" --filter-name "HighFS" \
    --filter-expression "MQ < 40.0" --filter-name "LowMQ" \
    -O ~/wes_analysis/results/variants/cohort_filtered.vcf.gz

Key Difference #4: Somatic Mutation Calling in WES

Somatic mutation detection follows the same Mutect2 approach as our tumor-normal mutation calling tutorial, but restricted to captured regions.

Tumor-Normal Mutation Calling

Creating Your Panel of Normals

# Define variables for somatic analysis
TUMOR_SAMPLE="tumor_sample_001"
NORMAL_SAMPLE="normal_sample_001"
GNOMAD="~/references/somatic_resources/af-only-gnomad.hg38.vcf.gz"

#=============================================================================
# CHOOSE YOUR PANEL OF NORMALS STRATEGY
#=============================================================================

# Option 1: Use Public PoN (RECOMMENDED for beginners and small studies)
# - Easier to implement
# - High quality from 1000+ samples
# - Good for most applications
PON="~/references/somatic_resources/1000g_pon.hg38.vcf.gz"

# Option 2: Create Custom PoN (ADVANCED - only if you have ≥40 normal samples)
# - Better for large research studies
# - Can be optimized for your specific capture kit
# - Requires significant computational resources
# See detailed workflow below if you choose this option

#=============================================================================
# ADVANCED: Custom PoN Creation (skip if using public PoN)
#=============================================================================

# Only proceed if you have ≥40 normal samples from the same capture kit
# Create output directory
mkdir -p ~/wes_analysis/results/somatic/pon_vcfs

# Step 1: Run Mutect2 on each normal sample in tumor-only mode
for NORMAL_BAM in ~/wes_analysis/results/aligned/normal*.bam; do
    NORMAL_ID=$(basename ${NORMAL_BAM} _recalibrated.bam)

    gatk Mutect2 \
        -R ${REFERENCE} \
        -I ${NORMAL_BAM} \
        -L ${TARGET_REGIONS} \
        --max-mnp-distance 0 \
        -O ~/wes_analysis/results/somatic/pon_vcfs/${NORMAL_ID}.vcf.gz
done

# Step 2: Create genomicsDB workspace
gatk GenomicsDBImport \
    -R ${REFERENCE} \
    -L ${TARGET_REGIONS} \
    --genomicsdb-workspace-path ~/wes_analysis/results/somatic/pon_db \
    --merge-input-intervals \
    -V ~/wes_analysis/results/somatic/pon_vcfs/*.vcf.gz

# Step 3: Create the final custom PoN
gatk CreateSomaticPanelOfNormals \
    -R ${REFERENCE} \
    -V gendb://~/wes_analysis/results/somatic/pon_db \
    -O ~/wes_analysis/results/somatic/custom_pon.vcf.gz

# Use custom PoN
PON="~/wes_analysis/results/somatic/custom_pon.vcf.gz"

Somatic Mutation Calling Using Mutect2

# Call somatic mutations with Mutect2
gatk Mutect2 \
    -R ${REFERENCE} \
    -I ~/wes_analysis/results/aligned/${TUMOR_SAMPLE}_recalibrated.bam \
    -I ~/wes_analysis/results/aligned/${NORMAL_SAMPLE}_recalibrated.bam \
    -normal ${NORMAL_SAMPLE} \
    --germline-resource ${GNOMAD} \
    --panel-of-normals ${PON} \
    -L ${TARGET_REGIONS} \
    -O ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_vs_${NORMAL_SAMPLE}_somatic.vcf.gz \
    -f1r2-tar-gz ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_f1r2.tar.gz

# Learn Read Orientation Model (filters orientation bias artifacts)
gatk LearnReadOrientationModel \
    -I ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_f1r2.tar.gz \
    -O ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_read_orientation_model.tar.gz

# Get Pileup Summaries for contamination estimation
gatk GetPileupSummaries \
    -I ~/wes_analysis/results/aligned/${TUMOR_SAMPLE}_recalibrated.bam \
    -V ${GNOMAD} \
    -L ${TARGET_REGIONS} \
    -O ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_pileups.table

gatk GetPileupSummaries \
    -I ~/wes_analysis/results/aligned/${NORMAL_SAMPLE}_recalibrated.bam \
    -V ${GNOMAD} \
    -L ${TARGET_REGIONS} \
    -O ~/wes_analysis/results/somatic/${NORMAL_SAMPLE}_pileups.table

# Calculate contamination
gatk CalculateContamination \
    -I ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_pileups.table \
    -matched ~/wes_analysis/results/somatic/${NORMAL_SAMPLE}_pileups.table \
    -O ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_contamination.table \
    --tumor-segmentation ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_segments.table

# Filter Mutect2 calls
gatk FilterMutectCalls \
    -R ${REFERENCE} \
    -V ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_vs_${NORMAL_SAMPLE}_somatic.vcf.gz \
    --contamination-table ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_contamination.table \
    --tumor-segmentation ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_segments.table \
    --ob-priors ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_read_orientation_model.tar.gz \
    -L ${TARGET_REGIONS} \
    -O ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_vs_${NORMAL_SAMPLE}_filtered.vcf.gz

# Extract PASS variants only
gatk SelectVariants \
    -R ${REFERENCE} \
    -V ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_vs_${NORMAL_SAMPLE}_filtered.vcf.gz \
    --exclude-filtered \
    -O ~/wes_analysis/results/somatic/${TUMOR_SAMPLE}_vs_${NORMAL_SAMPLE}_filtered_PASS.vcf.gz

Advantages of WES for Somatic Mutation Detection

Benefits of WES in cancer analysis:

Higher sensitivity for low-frequency mutations due to increased depth
Reduced false positives from repetitive/low-complexity regions
Faster analysis enables rapid clinical turnaround
Cost-effective for large cancer cohorts
Focused interpretation on actionable coding mutations

Limitations compared to WGS:

Misses regulatory mutations in promoters and enhancers
Cannot detect most structural variants
Limited CNV detection to exon-level changes
Splice site coverage may be incomplete

Variant Annotation: Same as WGS Analysis

Variant annotation for WES follows exactly the same process as our comprehensive Part 3: Annotating SNVs and Mutations with Multiple Tools tutorial.

The annotation tools (GATK Funcotator, Ensembl VEP, SnpEff, and ANNOVAR) work identically on VCF files regardless of whether the variants come from WGS or WES data. The key steps remain:

Functional Impact Prediction – Effect on protein coding sequences
Population Frequency Annotation – Adding gnomAD and 1000G frequencies
Clinical Significance – ClinVar pathogenicity classifications
Pathogenicity Scoring – SIFT, PolyPhen, CADD predictions

WES Annotation Advantages

Since WES focuses on protein-coding regions, annotation is often more straightforward:

Higher proportion of functionally relevant variants (most are in exons)
Better pathogenicity predictions (tools trained primarily on coding variants)
Clearer clinical interpretation (established for coding mutations)
Faster processing (fewer variants to annotate)

Recommendation: Follow Part 3 annotation tutorial using your filtered WES VCF files as input. All commands and databases work identically.

Key Difference #5: CNV Detection in WES Data

CNV detection from WES data requires specialized approaches due to the targeted nature of the sequencing and discrete coverage patterns.

Understanding WES CNV Challenges

Technical Challenges:

Capture bias: Different exons have vastly different capture efficiencies
Probe design: Some exons are difficult to capture due to sequence composition
Coverage gaps: Poorly captured exons create false CNV signals
Normalization complexity: Harder to distinguish true CNVs from technical artifacts

What WES CNV Analysis Can Detect:

Large gene deletions/duplications (>50kb, multiple exons)
Whole gene copy number changes
Multi-exon CNVs with clear coverage patterns
Dosage-sensitive gene alterations

CNV Detection Using CNVkit

# Create directory for CNV analysis
mkdir -p ~/wes_analysis/results/cnv
cd ~/wes_analysis/results/cnv

# Define sample paths
NORMAL_SAMPLES=(
    "~/wes_analysis/results/aligned/exome_normal1_recalibrated.bam"
    "~/wes_analysis/results/aligned/exome_normal2_recalibrated.bam"
    "~/wes_analysis/results/aligned/exome_normal3_recalibrated.bam"
)

TEST_SAMPLES=(
    "~/wes_analysis/results/aligned/exome_sample1_recalibrated.bam"
    "~/wes_analysis/results/aligned/exome_sample2_recalibrated.bam"
)

# Single command WES CNV analysis with target region specification
cnvkit.py batch \
    ${TEST_SAMPLES[@]} \
    --normal ${NORMAL_SAMPLES[@]} \
    --targets ${TARGET_REGIONS} \
    --fasta ${REFERENCE} \
    --annotate ~/references/cnv_references/refGene.bed \
    --output-reference wes_reference.cnn \
    --output-dir cnv_results/ \
    --method hybrid \
    --segment-method cbs \
    --drop-low-coverage \
    --scatter \
    --diagram

# Alternative: Auto-detect targets (when target file unavailable)
cnvkit.py batch \
    ${TEST_SAMPLES[@]} \
    --normal ${NORMAL_SAMPLES[@]} \
    --fasta ${REFERENCE} \
    --annotate ~/references/cnv_references/refGene.bed \
    --output-reference wes_auto_reference.cnn \
    --output-dir cnv_auto_results/ \
    --method hybrid \
    --target-avg-size 267 \
    --segment-method cbs \
    --scatter \
    --diagram

Alternative WES CNV Tools

ExomeDepth: Specifically designed for WES CNV detection

Accounts for exon-to-exon coverage variation
Uses matched reference samples from same capture kit
Lower false positive rate than general CNV tools
Widely used in clinical exome sequencing

CODEX: Another WES-specific CNV caller

Designed for whole exome and targeted sequencing
Uses negative binomial model for coverage
Good performance with small sample sizes

XHMM: Hidden Markov Model approach

Specifically for exome CNV detection
Good at handling coverage biases
Requires larger sample sizes for optimal performance

Clinical Interpretation: WES vs. WGS Considerations

Understanding WES Diagnostic Capabilities

What WES Can Diagnose:

Mendelian disorders caused by coding mutations (~85% of known disease variants)
Loss-of-function variants (nonsense, frameshift, splice site)
Missense variants in critical functional domains
Large CNVs affecting entire genes (with limitations)
Pharmacogenomic variants in drug metabolism genes

What WES Cannot Diagnose:

Regulatory variants in promoters, enhancers (~15% of disease variants)
Deep intronic variants affecting splicing
Structural variants not disrupting exons
Repeat expansions (Huntington’s, fragile X, etc.)
Balanced chromosomal rearrangements
Mosaic variants at low levels (<20-30%)

Best Practices for WES Analysis

Quality Control Standards

Coverage Requirements:

Mean target coverage: >100x for clinical applications, >50x for research
Coverage uniformity: >95% of targets at 20x depth
On-target rate: >80% for optimal results

Performance Benchmarks:

Variant calling sensitivity: >95% for variants >20x coverage
Specificity: >99% after proper filtering
Reproducibility: >98% concordance between technical replicates

Common Pitfalls and Solutions

Technical Issues:

1. Low On-Target Rate (<70%)

Causes: Wrong target file, poor capture, sample degradation
Solutions: Verify target file, check capture kit lot, assess DNA quality
Prevention: Use exact manufacturer target files, validate protocols

2. Uneven Coverage Across Targets

Causes: GC bias, capture probe efficiency, PCR artifacts
Solutions: GC bias correction, optimize PCR conditions
Monitoring: Track coverage uniformity metrics

3. High False Positive Rate

Causes: Insufficient filtering, capture artifacts, systematic biases
Solutions: Apply WES-specific filters, use population databases
Validation: Confirm suspicious variants with Sanger sequencing

Analytical Issues:

1. Missing Expected Variants

Causes: Low coverage in specific exons, variant in non-captured regions
Solutions: Manual review of coverage, consider WGS for negative cases
Prevention: Review capture kit content for genes of interest

2. CNV False Positives

Causes: Capture bias, reference sample selection, systematic artifacts
Solutions: Use WES-specific CNV tools, validate with orthogonal methods
Prevention: Include adequate normal samples in reference

3. Batch Effects

Causes: Different capture lots, library prep variations, sequencing platforms
Solutions: Batch correction methods, standardized protocols
Prevention: Process samples consistently, track batch information

4. Somatic Mutation Artifacts

Causes: DNA degradation, FFPE artifacts, low tumor purity
Solutions: Optimize filtering parameters, use orientation bias models
Prevention: Use high-quality samples, validate tumor content

Conclusion

Whole Exome Sequencing represents a powerful, cost-effective approach to genomic analysis that has revolutionized clinical genetics. While WES cannot replace WGS for comprehensive genomic analysis, it provides an excellent balance of coverage depth, cost efficiency, and clinical relevance for many applications.

Key Advantages of WES

Cost-effectiveness for large-scale studies and clinical applications
High coverage depth enabling detection of rare variants
Focused analysis on protein-coding regions with known disease relevance
Faster processing and interpretation compared to WGS
Established clinical workflows and interpretation guidelines

When to Consider WGS Instead

Previous WES analysis was negative but genetic cause still suspected
Regulatory variants or structural variants are clinically relevant
Comprehensive variant discovery is required for research
Non-coding regions need investigation for specific phenotypes

As sequencing costs continue to decrease and technologies advance, the choice between WES and WGS will increasingly depend on specific research questions, clinical requirements, and interpretation capabilities rather than cost alone. Both approaches remain valuable tools in the genomicist’s toolkit, each with distinct advantages for different applications.

References

Bamshad MJ, et al. (2011). Exome sequencing as a tool for Mendelian disease gene discovery. Nature Reviews Genetics, 12(11):745-55. doi:10.1038/nrg3031
Yang Y, et al. (2013). Clinical whole-exome sequencing for the diagnosis of mendelian disorders. New England Journal of Medicine, 369(16):1502-11. doi:10.1056/NEJMoa1306555
Retterer K, et al. (2016). Clinical application of whole-exome sequencing across clinical indications. Genetics in Medicine, 18(7):696-704. doi:10.1038/gim.2015.148
Chong JX, et al. (2015). The Genetic Basis of Mendelian Phenotypes: Discoveries, Challenges, and Opportunities. American Journal of Human Genetics, 97(2):199-215. doi:10.1016/j.ajhg.2015.06.009
Gilissen C, et al. (2012). Disease gene identification strategies for exome sequencing. European Journal of Human Genetics, 20(5):490-7. doi:10.1038/ejhg.2011.258
Rabbani B, et al. (2014). The promise of whole-exome sequencing in medical genetics. Journal of Human Genetics, 59(1):5-15. doi:10.1038/jhg.2013.114
MacArthur DG, et al. (2012). A systematic survey of loss-of-function variants in human protein-coding genes. Science, 335(6070):823-8. doi:10.1126/science.1215040
Lelieveld SH, et al. (2016). Meta-analysis of 2,104 trios provides support for 10 new genes for intellectual disability. Nature Neuroscience, 19(9):1194-6. doi:10.1038/nn.4352
Zook JM, et al. (2016). Extensive sequencing of seven human genomes to characterize benchmark reference materials. Scientific Data, 3:160025. doi:10.1038/sdata.2016.25
Richards S, et al. (2015). Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genetics in Medicine, 17(5):405-24. doi:10.1038/gim.2015.30
ATIA, T. (2019). Overview of genetic causes of recurrent miscarriage and the diagnostic approach. BIOCELL, 43(4), 253–262. https://doi.org/10.32604/biocell.2019.08180

This tutorial is part of the NGS101.com series on whole genome sequencing analysis. If this tutorial helped advance your research, please comment and share your experience to help other researchers! Subscribe to stay updated with our latest bioinformatics tutorials and resources.

Tags:

capture sequencing, CNV detection, exome vs genome, GATK Mutect2, mutation calling, targeted sequencing, variant calling, whole exome sequencing

Comments

14 responses to “How To Analyze Whole Exome Sequencing Data For Absolute Beginners: From Raw Reads to High-Quality Variants, Mutations, and CNVs”

Shivani Mishra

November 4, 2025

It really helped me a lot, please continjue making such content in bioinformatics.

Reply
1. Lei
  
  November 5, 2025
  
  I’m glad the tutorials have been helpful to you — more are on the way, so stay tuned!
  
  Reply
RITHIKA R S

November 4, 2025

Can you make a tutorial on how to do variant analysis and mutational studies using chip-seq data?
It will be of a great help if you do so. Hoping for a positive reply at the earliest convenience.

Thanks in advance.

Reply
1. Lei
  
  November 5, 2025
  
  I don’t think ChIP-seq is designed for variant analysis or mutational studies. May I ask where you came across such application?
  
  Reply
Charles

November 5, 2025

Thank your for these helpful tutorials. I was wondering if you would recommend these trainings for students who can only access windows PC. How should they go about it?

Reply
1. Lei
  
  November 5, 2025
  
  Hi Charles,
  
  I replied to your same question in the WGS tutorial part 1 (see below):
  
  Unfortunately, laptops lack the necessary power for whole-genome or whole-exome sequencing (WGS/WES) analysis. As noted in my tutorials, you’ll need access to a robust Linux system, such as an HPC cluster.
  If you’d like to gain hands-on experience with the full workflow, I’m launching a course that includes a pre-configured Linux environment. This will allow you to practice Linux command-line tools in a suitable setting. Let me know if you’re interested, or stay tuned for upcoming announcements!
  
  Reply
  1. Charles
    
    November 5, 2025
    
    Thank you for your prompt response. I’d like to gain hands-on-experience. I will stay tuned for a course that includes the pre-configured Linux environment.
    
    Reply
Najma

November 27, 2025

Hi Dr. Lei, thank you for this tutorial.
I have 10 WES samples and 2 controls. The controls are from healthy individuals, and the samples are from my diseased patients, but all are unrelated. So which mode should I use tumor-only or tumor-normal for somatic variant analysis? My disease is a chromosomal disorder, so how should I proceed? Do we need matched sample -control means diseased and control from same patient ?

Reply
1. Lei
  
  November 27, 2025
  
  Hi Najma,
  
  For the most reliable somatic mutation calling, it’s best to have matched tumor–normal (tumor + healthy tissue from the same patient) samples.
  
  If you only have unmatched tumor samples, you can follow my dedicated tutorial:
  How To Analyze Whole Genome Sequencing Data For Absolute Beginners – Part 2B: Unmatched Sample Mutation Calling Strategies
  
  If your samples are not from cancer and you instead want to identify patient- or disease-associated germline variants (or rare disease mutations), please use:
  How To Analyze Whole Genome Sequencing Data For Absolute Beginners – Part 5: Identifying Disease- or Patient-Specific Variants
  
  Reply
Najma

November 28, 2025

Thank you, Dr. Lei. I am currently working on a chromosomal disorder project. I was considering performing somatic variant calling as well, since I have already completed the germline analysis. However, somatic variant calling is not applicable to chromosomal disorders, correct?

Reply
1. Lei
  
  November 28, 2025
  
  MuTect2 is optimized for detecting somatic mutations in cancer by actively filtering out germline variants. In non-cancer diseases, the pathogenic variants are usually germline (inherited) rather than somatic. Even if a germline variant looks different between an affected sample and a control, MuTect2 will treat it as germline and not report it as a somatic mutation. Therefore, for non-cancer diseases, somatic mutation calling with MuTect2 is not recommended.
  
  Reply
Najma

December 1, 2025

Okay, thank you Dr. Lei.

Reply
Michelle

December 28, 2025

Has anyone heard of Varlociraptor for detecting somatic variants in tumor-only experiments?

Reply
1. Lei
  
  December 28, 2025
  
  Hi Michelle,
  
  Just to clarify: Varlociraptor is not a standalone variant caller and definetely not a replacement of Mutect2. It requires candidate variants as input (from tools like Mutect2) and then applies a Bayesian statistical model to compute posterior probabilities and enable precise FDR-controlled filtering.
  
  Mutect2 remains the gold standard for initial somatic variant detection, especially in tumor-normal pairs. In tumor-only mode, where false positives are a bigger concern, Varlociraptor serves as an excellent post-processing tool—it can refine calls from Mutect2 and filter them at your desired FDR level more effectively than hard thresholds alone.
  
  Reply