Whole exome sequencing (WES) selectively sequences the transcribed regions of the genome, providing a cost-effective alternative to whole genome sequencing. It produces a smaller, more manageable dataset for analysis. Key steps in WES data analysis include quality control of raw sequencing data, alignment of reads to a reference genome, variant calling, annotation of variants, and filtering of variants. Critical quality metrics are assessed at each step to ensure high quality data and accurate detection of variants.

1. Whole exome sequencing
(WES|WXS)
and its data analysis
Feb 28, 2023
Haibo Liu
Senior Bioinformatician
UMass Medical School, Worcester, MA
Email: haibol2017@gmail.com

2. Eukaryotic Exome
The human exome contains about 180,000 exons. These constitute about 1% of
the human genome (~40 Mb).

3. Exome sequencing
• A NGS method that selectively sequences the transcribed
regions of the genome.
• Provides a cost-effective alternative to WGS
• Produces a smaller, more manageable data set for faster, easier data
analysis (4–5 Gb WES vs ~90 Gb WGS)
• Identify both somatic and germline variants
• Single Nucleotide Polymorphisms (SNPs)
• Small Insertions-Deletions (indels)
• Loss of Heterozygosity (LOH)
• Copy Number Variants (CNVs), structural variants (SV)
• Microsatellite stability

4. Performance of WES in clinical studies

5. Workflow of WES

6. Genotyping by Microarray, WES, and WGS
(not updated, data
analysis cost not
included)

7. Experimental design of WES
• Tissue sampling
• Somatic mutations
• Tumor (tumor purity and freshness are critical)
• Normal tissue or blood sample
• Germline mutations
• Blood or any other tissue
• Sample size and sample population
• cohort (disease vs health)
• Trio, related family (non-carrier, carrier, and patient)
• Capture methods
• Sequencing strategies
• platform, PE|SE, UMI, read length, seq. depth

8. Rescue to DNA preparation from FFPE fixed
samples

9. Exome capture: Target-enrichment strategies
Array-based capture
https://en.wikipedia.org/wiki/Exome_sequencing
• Twist Exome 2.0 (Twist
Bioscience)
• Nextera Rapid Capture
Exomes (Illumina)
• xGen WES (IDT)
• SureSelect (Agilent)
• KAPA HyperExome
(Roche)
• SeqCap (NimblGen)
• …
Capture toolkits

10. UMI for detecting low frequency mutations for
prenatal or cancer research
The Cell3™ Target library preparation behind our whole exome enrichment incorporates error suppression
technology. This includes unique molecular indexes (UMIs) and unique dual indexes (UDIs), to remove both
PCR and sequencing errors and index hopping events. This error suppression technique, combined with our
excellent uniformity of coverage, allows you to confidently and accurately call mutations down to 0.1% VAF
and enables generation of sequencing libraries from as little as 1 ng cfDNA input.

11. Comparison of different library preparation methods

12. Comparison of different library preparation methods

13. Sequencing depth

14. Quality control in WES
Raw data QC BAM QC
variant QC

15. Raw data QC
• QC tools
• FastQC/MultiQC
• NGS QC toolkit (https://github.com/mjain-lab/NGSQCToolkit)
• QC-chain (contamination detection)
• PRINSEQ
• QC3
• Important QC metrics
• Base quality
• Nucleotide distribution along cycles
• GC content distribution
• Duplication rate
• Adaptor content
QC3

16. Read trimming
• Trimmomatic, cutadapt, fastp (auto adaptor detection), …
• Quality/adaptor trimming
• Don’t trim 5’ end (markduplicates)

17. From raw fastq to analysis-ready BAM

18. Aligner
• BWA-mem
• Bowtie2, Novoalign, GMAP

19. Selection of reference genomes
• Completeness
• Decoyed genome (1000 Genomes analysis pipeline)
• EBV (herpesvirus 4 type 1, AC:NC_007605) and decoy sequences
derived from HuRef, Human BAC and Fosmid clones and NA12878.
(~36Mb)
• T2T- CHM13v1.1, the latest, complete human reference
genome

20. Quality control in WES
Raw data QC BAM QC
variant QC

21. BAM QC
• Important QC metrics
• % of reads that map to the reference
• % of reads that map to the baits
• Coverage depth distribution (target regions)
• Coverage unevenness & Cohort Coverage Sparseness
• Insert size distribution
• Duplicate rate
• Tools
• Alfred
• QC3
• Various picard CollectMetrics tools
• covReport

22. Cohort Coverage Sparseness (CCS) and
Unevenness (UE) Scores for a detailed
assessment of the distribution of coverage of
sequence reads
https://www.nature.com/articles/s41598-017-01005-x

23. Local and global non-uniformity of
different capture toolkits

24. Differences from Capture toolkits

25. Differences from Capture toolkits
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4092227/

26. Exome probe design is one of the major
culprits
• Most of the observed bias in modern WES stems from
mappability limitations of short reads and exome probe design
rather than sequence composition.
https://www.nature.com/articles/s41598-020-59026-y

27. Alfred QC metrics
https://academic.oup.com/bioinformatics/article/35/14/2489/5232224
Alignment Metric DNA-Seq (WGS) DNA-Seq (Capture) RNA-Seq ChIP-Seq/ATAC-Seq Chart Type
Mapping Statistics ✔ ✔ ✔ ✔ Table
Duplicate Statistics ✔ ✔ ✔ ✔ Table
Sequencing Error Rates ✔ ✔ ✔ ✔ Table
Base Content Distribution ✔ ✔ ✔ ✔ Grouped Line Chart
Read Length Distribution ✔ ✔ ✔ ✔ Line Chart
Base Quality Distribution ✔ ✔ ✔ ✔ Line Chart
Coverage Histogram ✔ ✔ ✔ ✔ Line Chart
Insert Size Distribution ✔ ✔ ✔ ✔ Grouped Line Chart
InDel Size Distribution ✔ ✔ ✔ ✔ Grouped Line Chart
InDel Context ✔ ✔ ✔ ✔ Bar Chart
GC Content ✔ ✔ ✔ ✔ Grouped Line Chart
On-Target Rate ✔ Line Chart
Target Coverage Distribution ✔ Line Chart
TSS Enrichment ✔ Table
DNA pitch / Nucleosome pattern ✔ Grouped Line Chart
https://www.gear-genomics.com/docs/alfred/webapp/#featuresty-control)

28. CovReport

29. From BAM to VCF
GATK:
 Slop exon by 200 bp
 Analysis for each
chromosome

30. Variant callers
(Mutect2)
(HaplotypeCaller)
BreakSeq, LUMPY, Hydra,DELLY, CNVNator, Pindel
FreeBayes/SAMtools, DeepVariant

31. GATK Best practices for population-
based germline variant calling

32. GATK Mutect2 Best practices for population-
based soMATIC variant calling

33. Discrepancy of variants called by
different callers

34. Integrated variant calling
• Integration of multiple tools’ results
• Isma (integrative somatic mutation analysis)
• Ensemble Machine learning method
• BAYSIC
• SomaticSeq
• NeoMutate
• SMuRF
(Bartha and Gyorffy2019)
(Nanni et al. 2019)

35. Quality control in WES
Raw data QC BAM QC
variant QC

36. Sample-level Variant QC
• Tools
• GATK
CollectVariantCallingMetrics
, VCFtools, PLINK/seq, QC3
• Important QC metrics
• Ti/Tv ratio, nonsynonymous/synonymous,
heterozygous/nonreference-homozygous
(het/nonref-hom) ratio, mean depth,
• Genotype missing rate
• Genotype concordance to related data
(different platforms)
• Cross-sample DNA contamination
(VerifyBamID)
• Identity-by-descent (IBD) analysis (PLINK)
• Related samples
• PCA (EIGENSTRAT)
• Population stratum (ethnicity)
• Sex check (PLINK)

37. Ti/Tv ratio and het/nonref-hom ratio
• The Ti/Tv ratio varies greatly by genome region and
functionality, but not by ancestry.
• The het/nonref-hom ratio varies greatly by ancestry,
but not by genome regions and functionality.
• extreme guanine + cytosine content (either high or
low) is negatively associated with the Ti/Tv ratio
magnitude.
• when performing QC assessment using these two
measures, care must be taken to apply the correct
thresholds based on ancestry and genome region.
https://academic.oup.com/bioinformatics/article/31/3/318/2366248
Too low ==> high false positive rate; too high ==> bias.

38. Example report

39. Potential error sources in next-generation sequencing
workflow
https://genomebiology.biomedcentral.com/articles/10.1186/s13059-019-1659-6

40. Origin of variant artifacts
• Artifacts introduced by sample/library preparation
• low-quality base calls (Read-end artifacts and other low Qual bases)
• Alignment artifacts
• Local misalignment near indels,
• Erroneous alignments in low-complexity regions
• Paralogous alignments of reads not well represented in the reference
• Strand orientation bias artifacts (Strand Orientation Bias Detector
(SOBDetector), Fisher score)--
https://bmcgenomics.biomedcentral.com/articles/10.1186/1471-2164-13-666
•

41. Artifacts (https://genomemedicine.biomedcentral.com/articles/10.1186/s13073-020-00791-w)
Low base qual Read end
Strand bias Low complexity misalignment Paralog misalgnment

42. Variant-level QC
• Important QC metrics
• Genotype missing rate
• Hardy-Weinberg Equilibrium (caution) p-value
• Mendelian error rate
• Allele balance of heterozygous calls
• Variant quality score (GATK): filtering SNP and INDELS
separately(https://gatk.broadinstitute.org/hc/en-us/articles/360035890471-Hard-
filtering-germline-short-variants)
• Hard filter
• QualByDepth (QD)
• FisherStrand (FS)
• StrandOddsRatio (SOR)
• RMSMappingQuality (MQ)
• MappingQualityRankSumTest (MQRankSum)
• ReadPosRankSumTest (ReadPosRankSum)
• Machine learning-based filtering: Variant Quality Score Recalibration
--filterExpression "QD < 2.0 || FS > 60.0 || MQ < 40.0 || MQRankSum < -12.5 || ReadPosRankSum < -8.0"
--filterName "my_snp_filter"

43. Variant-level filtering
• Tools
• GATK, VCFtools, PLINK/Seq
• Sequencing data-based filtering
• Exclude potential artifacts
• Database-based filtering:
• Exclude known variants which are present in public SNP databases,
published studies or in-house databases as it is assumed that common
variants represent harmless variations
• Pedigree-based filtering
• Each generation introduces up to 4.5 deleterious mutations, it might be as
well that a de novo mutation is causing the disease.
• Function-based filtering
• Caution: risk removing the pathogenic variant

44. Allelic balance
https://www.cureffi.org/2012/09/19/exome-sequencing-pipeline-using-gatk/

45. Allelic balance
• SLIVAR: genotype quality, sequencing depth, allele balance, and
population allele frequency : https://github.com/brentp/slivar
https://onlinelibrary.wiley.com/doi/full/10.1002/humu.23674

46. Variant annotation tools
VAT Annotation of variants
by functionality in a
cloud computing
environment.

47. Variant annotation databases

48. Functional predictors/Prioritization tools
snpSift http://pcingola.github.io/SnpEff/ss_introduction/
(Hintzsche et al., 2016)
VAAST https://github.com/Yandell-Lab/VVP-pub
VarSifter, VarSight
gNome, KGGseq

49. (Cheng et al. Accurate proteome-wide missense variant effect prediction with AlphaMissense.
SCIENCE, 19 Sep 2023Vol 381, Issue 6664,DOI: 10.1126/science.adg7492)
Latest, advanced AI tool for infer effect of missense mutations: AlphaMissense
(Hintzsche et al., 2016)

50. Tools and resources for linking variants to
therapeutics

51. Variant visualization tools
VIVA, vcfR
oncoprint

52. Oncoprint for visualizing cohort variants

53. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6895801/
Beyond variants

54. Summary

55. WES and its data analysis

57. WES data analysis pipelines
• DRAGEN (Illumina)
• https://www.illumina.com/products/by-type/informatics-
products/basespace-sequence-hub/apps/dragen-enrichment.html
• JWES

58. • A high-performance commercial solution
(https://www.sentieon.com/products/)
• improves upon BWA, STAR, Minimap2, GATK, HaplotypeCaller,
Mutect, and Mutect2 based pipelines and is deployable on any
generic-CPU-based computing system
WES data analysis pipelines