Rare Variant Analysis Workflows: Analyzing NGS Data in Large Cohorts

Rare Variant Analysis
Workflows:
Analyzing NGS Data in
Large Cohorts
Nov 13, 2013
Bryce Christensen
Statistical Geneticist / Director of Services

Use the Questions pane in
your GoToWebinar window
Questions during
the presentation

Golden Helix
Leaders in Genetic Analytics
 Founded in 1998
 Multi-disciplinary: computer science,
bioinformatics, statistics, genetics
 Software and analytic services
About Golden Helix

GenomeBrowse
 Free sequencing visualization tool
 Launched in 2011
 Makes the process of exploring DNA-
seq and RNA-seq pile-up and
coverage data intuitive and powerful
 Stream public annotations via the
cloud
 Use it to validate variant calls, trio
exploration, de Novo discovery, and
more

Core
Features
Packages
Core Features
 Powerful Data Management
 Rich Visualizations
 Robust Statistics
 Flexible
 Easy-to-use
Applications
 Genotype Analysis
 DNA sequence analysis
 CNV Analysis
 RNA-seq differential expression
 Family Based Association
SNP & Variation Suite (SVS)

Previous Webcast Recording Available

Agenda
Brief review of upstream and QC considerations
NGS workflow design in SVS
2
3
4
Overview of RV analysis approaches
Define the problem: What is rare variant (RV) analysis?1
Interactive software demo5
What about exome chips?6
GenomeBrowse SVS 8: Exploratory tools, Analysis workflows

The Problem
 Array-based GWAS has been the primary technology for gene-
finding research for most of the past decade
- Common variant – common disease hypothesis
 NGS technology, particularly whole-exome sequencing, makes it
possible to include rare variants (RVs) in the analysis
 Individual RVs lack statistical power for standard GWAS
approaches
- How do we utilize that information?
 Proposed solution: combine RVs into logical groups and analyze
them as a single unit
- AKA “Collapsing” or “Burden” tests.

From the Vault:
January 2011 Slide on RV Analysis
What have we learned
since then?

NGS Analysis
Primary
Analysis
Secondary
Analysis
Tertiary
Analysis
“Sense Making”
 Analysis of hardware generated data, on-machine real-time stats.
 Production of sequence reads and quality scores
 Typical product is “FASTQ” file
 Recalibrating, de-duplication, QA and clipping/filtering reads
 Alignment/Assembly of reads
 Variant calling on aligned reads
 Typical products are “BAM” and/or “VCF” files
 QA and filtering of variant calls
 Annotation and filtering of variants
 Multi-sample integration
 Visualization of variants in genomic context
 Experiment-specific inheritance/population analysis
 “Small-N” and “Large-N” approaches

Most Importantly: Be Consistent!
Gholson Lyon, 2012

Things That Can Confound Your Experiment
Library preparation errors Sequencing errors Analysis errors
 PCR amplification point
mutations (e.g. TruSeq
protocol, amplicons)
 Emulsion PCR
amplification point
mutations (454, Ion
Torrent and SOLiD)
 Bridge amplification errors
(Illumina)
 Chimera generation
(particularly during
amplicon protocols)
 Sample contamination
 Amplification errors
associated with high or low
GC content
 PCR duplicates
 Base miscalls due to low
signal
 InDel errors (particular
PacBio)
 Short homopolymer
associated InDels (Ion
Torrent PGM)
 Post-homopolymeric tract
SNPs (Illumina) and/or
read-through problems
 Associated with inverted
repeats (Illumina)
 Specific motifs particularly
with older Illumina
chemistry
 Calling variants without
sufficient reads mapping
 Bad mapping (incorrectly
placed read)
 Correctly placed read
but InDels misaligned
 Multi-mapping to
paralogous regions
 Sequence contamination
e.g. adaptors
 Error in reference
sequence
 Alignment to ends of
contigs in draft assemblies
 Incorrect trimming of
reads, aligning adaptors
 Inclusion of PCR
duplicates
Nick Loman: Sequencing data: I want the truth! (You can’t handle the truth!)
Qual et al. A tale of three next generation sequencing platforms: comparison of Ion Torrent,
Pacific Biosciences and Illumina MiSeq sequencers. BMC Genomics. 2012 Jul

Two Primary Approaches
 Direct search for susceptibility variants
- Assume highly penetrant variant and/or Mendelian disease
- Extensive reliance on bioinformatics for variant annotation and filtering
- Sample sizes usually small—from single case up to nuclear families
 Rare Variant (RV) “collapsing” methods
- More common in complex disease research
- May require very large sample sizes!
- Assume that any of several LOF variants in a susceptibility gene may lead to
same disease or trait
- Many statistical tests available
- Also relies heavily on bioinformatics

Families of Collapsing Tests
 Burden Tests
- Combine minor alleles across multiple variant sites…
- Without weighting (CMC, CAST, CMAT)
- With fixed weights based on allele frequency (WSS, RWAS)
- With data-adaptive weights (Lin/Tang, KBAC)
- With data-adaptive thresholds (Step-Up, VT)
- With extensions to allow for effects in either direction (Ionita-Laza/Lange, C-alpha)
 Kernel Tests
- Allow for individual variant effects in either direction and permit covariate
adjustment based on kernel regression
- Kwee et al., AJHG, 2008
- SKAT
- SKAT-O
Credit: Schaid et al., Genet Epi, 2013

CMC: Combined Multivariate and Collapsing
 Multivariate test: simultaneous test for association of common
and rare variants in gene
 Flexibility in variant frequency bin definition
 Testing methods include Hotelling T2 and Regression
 Regression method allows for covariate correction
 Li and Leal, AJHG, 2008

KBAC: Kernel Based Adaptive Clustering
 Per-gene tests models the risk associated with multi-locus
genotypes at a per-gene level
 Adaptive weighting procedure that gives higher weights to
genotypes with higher sample risks
- Meant to attain good balance between classification accuracy and the number of
estimated parameters
 SVS implementation includes option for 1- or 2-tailed test
- But most powerful when all variants in gene have unidirectional effect
 Permutation testing or regression options
- Regression allows for covariate correction
 Liu and Leal, PLoS Genetics, 2010

NGS Analysis Workflow Development in SVS
 SVS is very flexible in workflow design.
 SVS includes a broad range of tools for data manipulation and
variant annotation and visualization that can be used together to
guide us on an interactive exploration of the data.
 We begin by defining the final goal and the steps needed to help us
reach that goal:
- Are we looking for a very rare, non-synonymous variant that causes a dominant
Mendelian trait?
- Are we looking for a gene with excess rare variation in cases vs controls?
 Once we know what we are looking for, we can identify the
available annotation sources that will help us answer the question.

Python Integration in SVS
 Allows rapid
development and
iteration of new
functions
 API access to most
SVS functions
 Access to extensive
Python analytic
libraries
 Fully documented in
manual

SVS Online Scripts Repository
 Downloadable add-on
functions for a variety of
analysis and data
management tasks
 “Plug-and-play”
 Some contributed by
customers
 Popular scripts often get
adopted into the “shipped”
version of SVS.
 Scripts in repository are
forward compatible to
SVS 8.0

Today’s Featured Scripts
 Activate Variants by Genotype Count Threshold
- Identify variants that occur with a specified frequency in one or several groups
 Filter by Marker Map Field
- Variant-level “INFO” fields from VCF files are imported to the SVS marker map
- This script allows you to filter markers based on those variables
 Many more useful scripts to take a look at:
- Add Annotation Data to Marker Map from Spreadsheet
- Nonparametric association tests
- Import Unsorted VCF Files
- Build Variant Spreadsheet
- Many, many more

Interactive Demonstration
 GenomeBrowse
- Exploring multi-sample VCF files in our free genome viewer software
 SVS 8.0
- Exploratory analysis workflow
- Using downloaded scripts
- Using basic analysis tools to create advanced workflows
- Simulate the development of a burden test
- RV association testing workflow
- KBAC
- CMC
- Data visualization

What about Exome Chips?
 Exome chips CAN be used
with RV association tests
 Exome chips include both
common and rare variants
 Remember: Exome chips
don’t capture all rare variants.
 Exome chips are thus less
powerful than WES for RV
associations, but also
significantly cheaper.

A Note about Exome Chips
 Exome chips are not GWAS chips
- GWAS chips focus on common SNPs, have uniform spacing, minimal LD and
are designed to capture population variability
- Exome chips include rare variants and the content is anything but uniform
 Most GWAS functions can be used with exome chips, but
require some workflow adjustments
- Gender checking
- IBD estimation
- Principal components
 Not unlike other chips with custom/targeted content
- Cardio-MetaboChip
- ImmunoChip

Questions or
more info:
 info@goldenhelix.com
 Request a copy of SVS at
www.goldenhelix.com
 Download GenomeBrowse
for free at
www.GenomeBrowse.com

Use the Questions pane in
your GoToWebinar window
Any Questions?

Rare Variant Analysis Workflows: Analyzing NGS Data in Large Cohorts

More Related Content

What's hot (20)

Similar to Rare Variant Analysis Workflows: Analyzing NGS Data in Large Cohorts (20)

More from Golden Helix Inc (20)

Recently uploaded (20)

Rare Variant Analysis Workflows: Analyzing NGS Data in Large Cohorts