Principles and Workflow of Whole Exome Sequencing: A Technical Guide for Project Design

Whole exome sequencing (WES) targets the protein-coding regions of the genome—approximately 35 Mb, or 1-2% of the total genome sequence—using hybridization-based capture enrichment combined with high-throughput sequencing. Despite covering only a small fraction of the genome, WES interrogates approximately 85% of known disease-causing mutations, making it one of the most cost-effective strategies for both research and clinical genomic analysis. The trade-off between genome coverage and sequencing cost—WES at 100× costs roughly one-fifth of WGS at 30×—positions exome sequencing as an accessible entry point for projects requiring comprehensive variant detection across many samples. All WES services and bioinformatics analysis described in this guide are for research use only and are not intended for clinical diagnostic applications.

This guide provides a technical framework for researchers designing WES projects. It covers the biochemical principles of hybrid capture enrichment that determine capture efficiency and uniformity, the quantitative relationship between sequencing depth and effective coverage, the key differences between tumor-normal paired and tumor-only analysis strategies, and the quality control metrics that distinguish high-quality WES data from compromised runs. The focus is on designing experiments that generate WES data with the depth and uniformity needed for the intended application—whether germline variant discovery, somatic mutation detection, or clinical diagnostic sequencing. Each section provides actionable guidance for a specific stage of the project design process, from capture kit selection through bioinformatics analysis to variant interpretation.

Whole exome sequencing services use validated capture kits and standardized library preparation protocols to achieve >95% of target bases at 20× coverage with fold-80 scores below 2.0. The choice of capture chemistry, sequencing platform, and depth directly determines the data quality and the types of variants that can be reliably detected—making informed project design the single most important factor in WES study success.

What Is Whole Exome Sequencing and When Should You Choose It?

Whole exome sequencing enriches and sequences the exonic regions of the genome—defined as the combined exon sequences of all protein-coding genes, plus untranslated regions (UTRs) and non-coding RNA genes (miRNA, lncRNA) included by most commercial capture kits. The human exome contains approximately 180,000 exons across 20,000-25,000 genes. While this represents only 1-2% of the genome, the exome harbors approximately 85% of known Mendelian disease-causing variants and a substantial fraction of cancer driver mutations, making it a highly efficient target for variant discovery.

The decision to use WES rather than WGS or a targeted gene panel depends on three factors: required coverage scope, sample number, and budget. WES provides genome-wide coding coverage at a cost per sample that enables analysis of hundreds to thousands of samples—a scale that would be prohibitively expensive with WGS. Targeted panels provide even higher depth at lower cost but are limited to predefined gene sets. For projects that require discovery of novel coding variants across many samples, WES provides the best balance of scope and cost efficiency. For projects focused on known genes or pathways, targeted panels are more cost-effective. For projects requiring comprehensive genome-wide analysis including non-coding regions, structural variants, and regulatory elements, WGS is required. Whole genome sequencing provides the most comprehensive view but at higher per-sample cost.

One additional factor often overlooked in the WES vs WGS decision is the analytical reproducibility across batches. WES data from different capture kits or different batches of the same kit show batch-specific coverage patterns that complicate cross-study comparisons and meta-analyses. WGS, because it does not depend on capture chemistry, provides more consistent coverage profiles across different laboratories and sequencing runs. This batch-effect consideration is relevant for large-scale multi-center studies or for projects that plan to integrate data from multiple sources.

Figure 1: WES vs WGS vs targeted panel — scope, depth, cost, and application fit

The Principle of Hybrid Capture — How Exome Enrichment Works

The core enabling technology of WES is hybridization-based capture enrichment, in which biotinylated DNA or RNA probes (baits) complementary to exonic sequences are hybridized to fragmented genomic DNA, captured on streptavidin-coated magnetic beads, and washed to remove unbound non-target DNA. Understanding the design parameters and limitations of this process is essential for interpreting WES data quality.

Probe design parameters: Commercial exome capture kits use probes of 60-120 nucleotides, designed with a tiling density that covers each target region with 2× overlapping probes—meaning every target base is covered by at least two independent probes from different positions. This tiling strategy ensures that if one probe in a region fails to capture its target due to sequence variation or secondary structure, the overlapping probe provides redundant coverage. The probe set for a typical human exome capture kit contains 400,000-700,000 unique probes, depending on the target region design and tiling density. A 2025 algorithm published in Bioinformatics (OLTA) optimizes bait selection to minimize the number of probes needed while maintaining target coverage, reducing capture costs without sacrificing efficiency.

Hybridization conditions: Hybridization is performed at 65°C for 16-24 hours in the standard protocol, or at elevated temperatures with shorter times (1.5-4 hours) in rapid capture protocols that use higher probe concentrations and optimized buffer formulations. Stringency washes at 65°C with decreasing salt concentrations remove partially hybridized non-target DNA. The stringency of these washes directly determines the on-target rate—more stringent washes increase the proportion of reads mapping to target regions but reduce overall yield, while less stringent washes capture more off-target DNA (including non-target genomic regions and mitochondrial DNA) that can be informative for CNV analysis but reduces effective sequencing depth on target. The hybridization temperature must be precisely controlled—every 1°C deviation from the optimal temperature reduces capture efficiency by approximately 5-10% for GC-matched targets and more for AT-rich or GC-rich regions.

Fold-80: the most important uniformity metric: Fold-80 measures how many times the average depth must be sequenced to achieve 80% of target bases at that average depth. A fold-80 of 1.0 represents perfect uniformity (all target bases at the same depth). A fold-80 of 2.0 means 1.8× the average depth is needed to cover 80% of targets at that depth. The Twist Bioscience white paper demonstrated that improvements to fold-80 (uniformity) have a substantially larger impact on effective coverage than improvements to on-target rate. For example, improving fold-80 from 2.0 to 1.5 reduces the sequencing required to achieve 80% of bases at 20× by approximately 25%, while an equivalent improvement in on-target rate (from 70% to 80%) reduces required sequencing by only 12%. This makes fold-80 the most actionable metric for evaluating WES data quality and for designing sequencing budgets.

2024 kit benchmark data: A 2024 comparative evaluation of four exome enrichment solutions (Roche, Agilent, Vazyme, Nanodigmbio) published in BMC Genomics in 2025 provides the most recent head-to-head performance data. All four kits achieved >97.5% of target bases at 10× and >95% at 20×. Roche exhibited the most uniform coverage (lowest fold-80), while Nanodigmbio achieved the highest on-target rate due to fewer off-target reads. Variant detection sensitivity was high across all kits for SNVs (>99% at 20×) but varied more for indels (95-98%), where capture uniformity had the greatest impact. The key finding for project design is that kit selection should prioritize uniformity metrics (fold-80) over on-target rate when the research goal is comprehensive variant detection, particularly for clinically relevant regions where coverage failure at individual exons can mean missing a pathogenic variant. For projects focused on a specific gene panel or known genomic regions, kits with targeted optimization for those regions may outperform general-purpose exome kits.

Practical considerations for capture kit selection: Beyond fold-80 and on-target rates, kit selection should account for the target region definition. Some kits include flanking intronic regions essential for splice site analysis, UTRs important for regulatory variant detection, and non-coding RNA genes relevant to specific diseases. The CCDS (Consensus CDS) coverage—the fraction of well-annotated coding exons included in the kit's target region—varies between 92-98% across commercial kits. For projects focused on disease gene discovery, a kit with higher CCDS coverage may be preferred even if its on-target rate is slightly lower. For large cohort studies where per-sample cost is the primary constraint, the kit with the most efficient capture (highest on-target rate with acceptable fold-80) provides the best value.

Figure 2: Complete WES workflow — from sample preparation through bioinformatics analysis

WES vs WGS vs Targeted Panel — Three Approaches Compared

The three sequencing strategies differ in genome coverage, sequencing depth, cost per sample, and the types of variants they can reliably detect.

WGS at 30× covers the entire genome including coding, non-coding, and regulatory regions. It detects SNVs, small indels, structural variants, and CNVs across the full genome, with no capture bias and no regions systematically excluded by probe design. However, at 30×, the average depth on coding regions is lower than WES, reducing sensitivity for low-frequency somatic variants. At scale, WGS is the most expensive of the three approaches for coding-targeted analysis.

WES at 100-200× covers only the target regions of the capture kit (typically 35-50 Mb of exonic and flanking sequence). The higher depth enables detection of low-frequency somatic variants down to 5-10% allele frequency in high-purity samples. WES detects coding SNVs and small indels with high sensitivity but has limited capability for CNV detection in captured regions and no ability to detect variants in non-targeted regions. For large-scale coding-focused variant discovery across hundreds of samples, WES provides the most efficient balance of sequencing cost and data comprehensiveness.

Targeted panels cover 0.1-5 Mb of selected genes or regions at very high depth (500-2,000×). This depth enables detection of somatic variants at 1-5% allele frequency, making panels the method of choice for liquid biopsy assays and for monitoring minimal residual disease. Panels are the most cost-effective approach for predefined gene sets but offer no discovery potential beyond the targeted genes.

For most research projects, the choice between the three follows a clear logic: use WGS when the budget allows and the question requires genome-wide analysis. Use WES when the budget constrains sample throughput and the question focuses on coding variants. Use targeted panels when the genes of interest are well-defined and maximum depth is required. An increasingly common strategy is a staged approach—starting with WES for broad variant discovery across a cohort, then validating and following up specific findings with targeted deep sequencing or functional studies. This design maximizes the discovery potential of WES while controlling the overall project cost by reserving more expensive assays for the validation phase.

Coverage Depth Requirements — How Much Sequencing Is Enough?

The required sequencing depth for WES depends on the variant types to be detected and the expected allele frequency. For germline SNV detection, 100× mean target coverage provides >99% sensitivity for heterozygous variants at allele frequency 0.5. The critical QC metric is the fraction of target bases covered at sufficient depth—for germline diagnostics, >95% of target bases at 20× is the minimum standard, and >98% at 20× is the target for clinical-grade data.

For somatic variant detection in cancer, higher depth is required because somatic mutations are present at lower allele frequencies depending on tumor purity. At 150-200× mean target coverage, WES detects somatic SNVs at 10-20% allele frequency with reasonable sensitivity. For lower-frequency variants, deeper sequencing or targeted approaches are needed. The relationship between mean depth, allele frequency, and detection sensitivity follows a Poisson distribution—to detect a variant at 5% allele frequency with 95% confidence requires at least 200× coverage at the variant position.

The uniformity of coverage (fold-80) directly affects how much sequencing is needed to achieve these depth metrics. A library with fold-80 of 1.6 requires 40% less sequencing than one with fold-80 of 2.4 to achieve the same percentage of target bases at 20×. This makes fold-80 optimization one of the most cost-effective strategies for improving WES project efficiency. Coverage requirements for WES projects provides a detailed guide to depth planning for different applications.

Practical depth planning: The relationship between mean target depth, sample number, and sequencing cost follows a straightforward calculation. A standard exome capture kit targets 35-50 Mb. At 100× mean depth, this requires 3.5-5 Gb of on-target reads. With on-target rates of 60-75%, the total sequencing required is approximately 5-8 Gb per sample. For a 96-sample project multiplexed on a NovaSeq 6000 S4 flow cell producing 1,000 Gb of data, this translates to approximately 120-190 samples per flow cell, depending on on-target efficiency. For LP-WES applications where 30-50× is sufficient for population screening, the per-sample cost can be reduced by 50-60%. These calculations should be validated against the actual performance of the selected capture kit and sequencing platform before committing to project-scale budgeting.

Figure 3: Coverage depth vs effective coverage — the impact of fold-80 on sequencing efficiency

WES Bioinformatics Pipeline — Key Adjustments for Exome Data

The standard GATK Best Practices pipeline for variant discovery requires specific adjustments for WES data that differ from WGS analysis. The non-uniform coverage of capture-based WES introduces systematic bias that affects multiple steps in the pipeline.

Pre-processing adjustments: Unlike WGS, where duplicate marking is primarily for PCR duplicate removal, WES data requires careful duplicate handling because the capture process itself produces a higher proportion of duplicate reads from the same original DNA molecule. Using Picard MarkDuplicates with the REMOVE_DUPLICATES=false option (marking but not removing) allows downstream tools to handle duplicates appropriately. Pre-indexing of the deduplicated BAM file is essential for GATK processing.

Variant calling in WES data: GATK HaplotypeCaller should be run with the --exome-mode flag for WES data, which adjusts the active region detection to account for the non-contiguous coverage of exome targets. For somatic variant calling in tumor-normal paired WES data, Mutect2 is the recommended caller. A 2025 benchmark in MDPI Biomolecules comparing Mutect2, Strelka2, and FreeBayes for somatic WES found that Mutect2 achieved the highest precision (>95%) while Strelka2 achieved the highest recall (>90%) for SNVs. For tumor-only WES analysis (where no matched normal is available), sensitivity decreases by 15-20% compared to paired analysis because germline variants cannot be distinguished from somatic mutations. Tumor-only WES is appropriate for initial screening but paired analysis is strongly recommended for studies requiring accurate somatic variant identification.

Filtering and annotation adjustments for WES: The lower number of total reads in WES compared to WGS means that variant filtering thresholds should be adjusted accordingly. For germline WES, a minimum depth of 10× and a minimum alternative allele count of 3 are standard filtering criteria. For somatic WES, filtering at 20× with 5 supporting reads reduces false positives caused by capture artifacts. Annotation pipelines for WES should include gene-level coverage flags indicating which exons had insufficient depth for reliable variant calling — without these flags, a lack of variant calls in a poorly covered exon could be incorrectly interpreted as the absence of a mutation. The VEP (Variant Effect Predictor) or SnpEff annotation tools generate these coverage annotations when provided with the WES target region BED file.

CNV detection from WES data: CNV detection from WES is fundamentally more challenging than from WGS because the capture process introduces non-uniform coverage that varies between samples and between capture batches. The standard approach uses a pooled reference of ≥30 matched normal samples to model the capture-specific noise profile, then compares each sample's coverage to the reference. ECOLE (2023, Nature Communications), a deep-learning-based CNV caller, is specifically designed for WES data and achieves 20-30% fewer false positives than conventional methods.

Turnaround time and computational resources: Standard WES bioinformatics processing for a 96-sample project—from FASTQ alignment through variant calling, annotation, and QC reporting—requires approximately 8-16 hours on a compute cluster with 32+ CPU cores, or 24-48 hours on a high-end workstation with 16 cores and 64 GB RAM. Storage requirements are approximately 2-5 GB per sample for aligned BAM files (compressed) and 200-500 MB per sample for VCF files and analysis outputs. For projects with >500 samples, cloud-based processing or a local compute cluster is recommended to manage run times.

Figure 4: WES application areas — inherited disease, cancer genomics, and liquid biopsy

Applications of WES in Research and Clinical Genomics

Inherited disease diagnosis: Trio-based WES (sequencing the proband and both parents) achieves diagnostic rates of 25-50% for suspected genetic disorders, with the highest yields in neurodevelopmental disorders, epilepsy, and congenital anomalies. The discovery rate for novel disease-gene associations has accelerated with large-scale WES cohorts such as the 100,000 Genomes Project and the Undiagnosed Diseases Network. The analytical framework for diagnostic WES follows a systematic filtering pipeline: rare variants (population frequency < 0.1%) are prioritized, followed by assessment of predicted functional impact (nonsense, frameshift, splice site), inheritance pattern compatibility, and gene-level phenotype matching. For proband-only WES, the diagnostic rate drops to 15-30% due to the inability to filter by inheritance, but this approach remains common in research settings where parental samples are unavailable. Population-scale WES projects such as UK Biobank's 200,000 exome dataset are enabling gene-level burden testing that identifies novel disease associations by aggregating rare variants within genes across large cohorts.

Cancer genomics: Tumor-normal paired WES is the standard approach for identifying somatic driver mutations, calculating tumor mutational burden (TMB), and detecting mutational signatures. For homologous recombination deficiency (HRD) assessment—a predictive biomarker for PARP inhibitor therapy—WES-based HRD scores derived from genome-wide loss of heterozygosity patterns have been validated in multiple cancer types. Cancer WES services include tumor-normal paired sequencing with bioinformatics analysis for somatic SNV, indel, and CNV detection.

Liquid biopsy WES: A 2025 study in Nature Scientific Reports validated an AI-enabled exome/transcriptome liquid biopsy assay (Caris Assure) that couples WES of circulating tumor DNA with machine learning for multi-cancer detection. While WES for ctDNA faces the challenge of low tumor DNA fraction in plasma (often < 1%), recent advances in hybrid capture chemistry and computational deconvolution are making cfDNA WES increasingly viable for non-invasive cancer profiling. For early-stage cancers where ctDNA fraction is lowest, WES-based liquid biopsy currently requires higher depths (500-1,000× on target) to detect the small number of tumor-derived fragments in the cfDNA pool. The key innovation enabling cfDNA WES is the use of unique molecular identifiers (UMIs) to collapse PCR duplicates into consensus sequences, dramatically reducing the noise floor and enabling accurate variant detection from as few as 10-100 template molecules per target region.

WES Data Analysis and Interpretation — From Variant Lists to Biological Insight

The output of a WES bioinformatics pipeline is a list of variants annotated with their genomic position, allele frequency, functional impact, and population frequency. Converting this variant list into interpretable biological or clinical findings requires a systematic filtering and prioritization framework.

Germline variant interpretation: For inherited disease studies, the filtering cascade starts with removing variants above 1% population frequency in gnomAD or ExAC, retaining only rare or novel variants. Next, variants are classified by predicted functional impact: protein-truncating variants (nonsense, frameshift, essential splice site) are prioritized over missense variants, which require additional evidence from conservation scores (PhyloP, GERP) and in silico prediction tools (SIFT, PolyPhen-2, CADD). The remaining candidate variants are evaluated for inheritance pattern compatibility (autosomal dominant, recessive, X-linked, de novo) and for concordance with the phenotype using tools such as Exomiser or Phen2Gene. Variant interpretation services provide systematic filtering pipelines that integrate population databases, functional predictions, and phenotype matching.

Somatic variant interpretation: Cancer WES analysis prioritizes variants by their recurrence across samples within a tumor type, their presence in the COSMIC cancer gene census, and their predicted impact on protein function. Tumor mutational burden (TMB) is calculated as the number of somatic coding mutations per megabase of genome sequenced. TMB derived from WES correlates well with WGS-based TMB and is used as a predictive biomarker for immunotherapy response. Mutational signature analysis—decomposing the spectrum of somatic mutations into characteristic patterns associated with specific mutational processes (e.g., APOBEC activity, smoking signature, UV damage)—is performed using tools such as SigProfiler or MutationalPatterns and requires at least 50-100 somatic mutations genome-wide for reliable signature assignment.

Reporting considerations: WES analysis reports should document not only the variants found but also the coverage performance across the target region. A common reporting standard includes the number of exon targets with insufficient coverage for variant calling (<10× for germline, <20× for somatic), the fold-80 uniformity metric, and the overall on-target rate. Variants reported from poorly covered exons should be flagged as low-confidence in the output. For clinical WES reporting, the ACMG guidelines for variant classification (pathogenic, likely pathogenic, VUS, likely benign, benign) should be followed, with supporting evidence documented for each classification tier.

Key Technical Challenges in WES Projects

Inter-batch capture variation: Hybrid capture efficiency varies between batches due to differences in reagent lot, hybridization temperature profiles, and operator technique. This batch effect introduces systematic differences in coverage uniformity that can mimic biological variation in downstream analyses. The 2024 BMC Genomics benchmark found that inter-batch variation accounted for 5-15% of coverage variability at individual exons, even within the same kit. Standard practice is to process all samples in a project within as few capture batches as possible, to capture batch-balanced controls, and to include replicate samples across batches for technical variation assessment.

GC bias: High-GC regions (>65% GC), which include many promoter regions and first exons of housekeeping genes, are significantly under-represented in WES data because GC-rich DNA has lower hybridization efficiency and higher secondary structure that impedes capture. GC-biased baits that include degenerate nucleotides or modified bases can partially compensate, but some GC-rich exons consistently fail to achieve target coverage across all commercial kits. These systematic coverage gaps should be identified and documented in the analysis methods, as variants in these regions cannot be reliably assessed from WES data.

FFPE artifacts: FFPE samples have fragmented DNA (average size < 300 bp) and deaminated bases from formalin cross-linking. The fragmentation reduces capture efficiency because shorter fragments hybridize less stably to baits—a 150 bp fragment has approximately 70% of the capture efficiency of a 300 bp fragment under standard hybridization conditions. Increasing DNA input (200-500 ng for FFPE vs 50-100 ng for fresh tissue) and reducing the hybridization temperature to 60°C partially compensates, but FFPE WES data consistently has lower fold-80 uniformity and higher duplicate rates than fresh-frozen data. For projects involving archived FFPE samples, validating the capture efficiency on a test sample before committing to full-scale library preparation can identify whether the DNA quality is sufficient for the required coverage metrics.

PCR duplicate accumulation: WES libraries have inherently higher PCR duplicate rates than WGS libraries because the capture step concentrates sequencing capacity onto a small target region, amplifying any library amplification bias. Duplicate rates above 25% indicate that the library complexity—the number of unique DNA fragments available for sequencing—is insufficient for the target depth. This typically results from low DNA input (<50 ng for standard protocols) or from over-cycling during the pre-capture PCR amplification. Reducing PCR cycles from 14-16 to 10-12 in the pre-capture PCR step, combined with using 100-200 ng of input DNA, can reduce duplicate rates below 15% while maintaining sufficient library yield for capture.

QC Metrics for WES Data

The QC metrics table serves as a practical checklist for evaluating WES data before proceeding with downstream analysis. A sample that falls below the minimum acceptable threshold for any of these metrics should be flagged for potential repeat sequencing or exclusion from the analysis. In large cohort studies, it is common to reject 3-5% of samples at the QC stage, and this expected rejection rate should be factored into the project sample size planning to ensure sufficient sample numbers after QC filtering.

Figure 5: WES data analysis and filtering pipeline — from raw reads to annotated variants

Figure 6: WES data quality assessment pyramid — hierarchy of QC metrics from sample input to variant output

FAQ

What sequencing depth do I need for WES?
For germline SNV detection, 100× mean target coverage is standard. For somatic variant detection in cancer, 150-200× is recommended. The key metric is the fraction of target bases at sufficient depth—>95% of targets at 20× for germline analysis.

What is fold-80 and why does it matter?
Fold-80 measures coverage uniformity across target regions. A fold-80 of 1.0 means perfect uniformity; lower fold-80 means less sequencing is needed to cover all target exons at required depth. It is the most actionable metric for evaluating WES capture efficiency.

Can WES detect CNVs?
Yes, but with lower sensitivity than WGS due to the non-uniform coverage of capture-based enrichment. CNV detection from WES requires a reference set of ≥30 normal samples and tools specifically designed for WES data, such as ECOLE or CNVkit with appropriate reference construction.

How does FFPE sample quality affect WES results?
FFPE samples produce fragmented DNA that reduces capture efficiency and coverage uniformity. Fold-80 typically increases by 20-30% for FFPE compared to fresh-frozen samples. Increasing DNA input and using FFPE-optimized capture protocols can partially mitigate this.

Should I use tumor-only or tumor-normal paired WES for cancer analysis?
Tumor-normal paired WES enables distinction between somatic mutations and inherited germline variants, reducing false-positive rates and enabling detection of clonal hematopoiesis artifacts. Tumor-only WES has 15-20% lower sensitivity for somatic variant detection but can be used for screening when matched normal tissue is unavailable.

What is the diagnostic yield of WES for genetic disorders?
Trio-based WES achieves diagnostic rates of 25-50% for suspected genetic disorders, with the highest yields in neurodevelopmental disorders and congenital anomalies. Proband-only WES has lower diagnostic rates (15-30%).

References

1. Comparative evaluation of four exome enrichment solutions in 2024. BMC Genomics. 2025;26:11196.
2. Methods, applications, and computational challenges in bait capture enrichment. Cell Reports Methods. 2025;5:100210.
3. OLTA: Optimizing bait selection for targeted sequencing. Bioinformatics. 2025;41:btaf146.
4. Comparative evaluation of Mutect2, Strelka2, and FreeBayes for somatic variant detection from WES. Biomolecules. 2025;15:1532.
5. Validation of an AI-enabled exome/transcriptome liquid biopsy assay. Nature Scientific Reports. 2025;15:8986.
6. ECOLE: Learning to call copy number variants on WES data. Nature Communications. 2023;14:44116.

For research purposes only, not intended for clinical diagnosis, treatment, or individual health assessments.