Rare Variant, Burden and Gene-Based Association Analysis for Cohorts

Overview

Rare variants are observed too infrequently for many single-variant association tests to achieve useful power. Rare variant association analysis addresses this problem by grouping variants within a gene, transcript, functional element, pathway, or predefined region and testing their combined relationship with a phenotype. The result is a set-based signal that can reveal aggregated genetic effects while reducing the number of tests relative to evaluating every rare variant separately.

CD Genomics supports WES and WGS cohorts, as well as external VCF, phenotype, and covariate data, with cohort QC, variant annotation, frequency and functional filtering, grouping, burden tests, SKAT, SKAT-O, covariate adjustment, population-structure or relatedness handling, multiple-testing control, visualization, and candidate prioritization. Analysis choices are documented because no single test is uniformly most powerful across all genetic architectures.

Aggregate sparse evidence: combine eligible rare variants into biologically meaningful sets.
Match tests to effect architecture: compare burden, variance-component, and omnibus approaches rather than assuming all variants act in one direction.
Control cohort confounding: align phenotype, ancestry, relatedness, technical QC, and variant definitions before testing.

Rare variants grouped by genes or regions for burden, SKAT, SKAT-O, covariate-adjusted testing, and candidate prioritization. Figure 1: Rare variant association analysis converts sparse variant observations into predefined set-level tests.

Why Single-Variant Tests Often Lose Power

A rare allele may occur in only a few participants, leaving too little information for a stable single-site estimate. Aggregation increases the number of informative carriers by testing multiple qualifying variants together. The gain depends on the grouping rule, allele-frequency threshold, functional annotation, causal-variant proportion, effect directions, sample size, phenotype distribution, and ancestry composition.

Variant sets should be defined before association results are inspected. Common units include genes, exons, protein domains, enhancers, sliding windows, and pathways. Multiple masks may capture different biological assumptions—for example, protein-truncating variants alone versus protein-truncating plus predicted damaging missense variants—but every additional mask increases the testing burden and should have a stated rationale.

Burden, SKAT, and SKAT-O

Method	Working Assumption	When It Can Be Useful
Burden test	Eligible variants influence the phenotype in a broadly consistent direction; collapsing or weighting creates one burden variable per set.	Sets enriched for likely functional variants with a relatively high causal proportion and aligned effects.
SKAT	Variant effects may differ in magnitude or direction, and some eligible variants may be neutral.	Heterogeneous sets where protective and risk effects may coexist or only a subset of variants is causal.
SKAT-O	The optimal balance between burden-like and variance-component evidence is not known beforehand.	Broad scans where genetic architecture varies among genes and an adaptive omnibus test is appropriate.

The test is only one component of the analysis. Variant selection, weights, null model, phenotype type, covariates, relatedness, case-control imbalance, and sample size can change calibration and power. Sensitivity analyses across justified masks or frequency thresholds can show whether a signal is robust or driven by one modeling choice.

Decision diagram comparing burden tests, SKAT, and SKAT-O under aligned, mixed, or uncertain rare variant effect patterns. Figure 2: Test selection follows the expected proportion and direction of causal effects, while SKAT-O adapts across a range of architectures.

Inputs and Cohort Requirements

Genotypes: jointly processed WES/WGS variants or a documented cohort VCF with genotype-level quality fields.
Phenotypes: clearly coded binary or quantitative traits and, where supported, additional outcome types agreed during scoping.
Covariates: age, sex, study site, batch, ancestry components, exposures, and other prespecified variables relevant to the model.
Sample metadata: sample IDs, inclusion/exclusion status, family or relatedness information, and consent-compatible population labels.
Annotation resources: reference assembly, transcript set, gene boundaries, consequence definitions, population frequencies, and functional scores.

When sequencing is required, related entry points include Whole Exome Sequencing and Whole Genome Resequencing. Existing sequence data can first undergo Variant Calling and cohort harmonization.

Which Input Path Fits Your Project?

Project Situation	Recommended Path	Key Consideration
Jointly processed WES/WGS variants with phenotype and covariate data	Direct entry: annotate, filter, construct sets, and test	Genotype quality, ancestry, and callability are verified before testing.
Existing cohort VCF with documented provenance	Variant QC and harmonization before association testing	Cross-processed cases and controls may need recalling or harmonization.
Raw FASTQ or BAM files only	Variant Calling followed by this analysis	Consistent calling across the cohort is required for set-based tests.
Sequencing not yet performed	Whole Exome Sequencing or Whole Genome Resequencing then analysis	Depth, cohort size, and study design should be aligned with the planned tests.
Small cohort or very few expected carriers	Power and expected-carrier evaluation before analysis	Sparse data may require SKAT-O or simulation-based feasibility review.
Phenotype, covariates, or relatedness information incomplete	Complete metadata before starting	Underspecified phenotypes and missing covariates degrade every downstream test.

Analysis Workflow

1. Analysis specification

We define the phenotype, cohort, hypothesis, grouping units, MAF thresholds, functional masks, covariates, relatedness strategy, tests, and multiple-testing plan before association analysis.

2. Sample and genotype QC

Sample identity, sex checks when appropriate, duplicates, missingness, depth or genotype quality, ancestry, relatedness, and batch are reviewed. Variant QC is aligned across groups.

3. Annotation and rare-variant filtering

Variants are annotated against the agreed assembly and transcript set, then filtered using cohort and external frequencies, consequence classes, callability, and project-specific evidence.

4. Set and mask construction

Eligible variants are grouped into genes or other predefined regions. Set sizes, carrier counts, and variants contributing to each mask are recorded.

5. Null model and association testing

Phenotype-appropriate models incorporate covariates, ancestry, and relatedness as required. Burden, SKAT, SKAT-O, or agreed methods are applied with calibrated handling of cohort structure.

6. Multiple testing and sensitivity review

Family-wise or false-discovery procedures are applied to the defined test family. Signals are compared across masks, methods, ancestry groups, or leave-one-variant-out checks when justified.

7. Prioritization and delivery

Significant and suggestive sets are linked to contributing variants, annotations, effect summaries, QC, plots, and a methods report for downstream research.

End-to-end rare variant analysis workflow from analysis specification and QC through annotation, mask construction, association testing, sensitivity review, and prioritization. Figure 3: Rare variant analysis is specified, tested, and reviewed as one connected set-based workflow rather than a single test.

Rare Variant Analysis Project Specifications

Parameter	Typical Scope	Review Point	Why It Matters
Genotype input	Jointly processed WES/WGS variants or a documented cohort VCF	Genotype-level quality fields, reference assembly, and callability	Comparable variants across the cohort are the foundation of set tests.
Phenotype and covariates	Coded binary or quantitative traits; age, sex, site, batch, ancestry components	Phenotype definition, missingness, and case-control balance	Underspecified phenotypes and omitted covariates distort calibration.
Variant filtering	MAF thresholds, consequence classes, and functional scores	Cohort and external frequencies, callability, and evidence sources	Prespecified thresholds keep filtering rules transparent.
Set and mask construction	Gene, region, or pathway units with one or more justified masks	Set sizes, carrier counts, and stated rationale per mask	Every additional mask increases testing burden and needs a reason.
Testing strategy	Burden, SKAT, SKAT-O, or agreed methods	Null model, covariates, relatedness, and ancestry handling	No single test is uniformly most powerful across genetic architectures.
Multiple testing and sensitivity	Family-wise or false-discovery procedures; mask and method comparisons	Leave-one-variant-out, ancestry strata, and convergence checks	Sensitivity review shows whether a signal depends on one modeling choice.
Deliverables	QC summaries, annotated variant table, set definitions, result tables, figures, prioritized candidates, report	Effect, p-value, multiple-testing, carrier-count, and convergence fields	Traceable candidates let downstream research verify every association.

Population Structure, Relatedness, and Imbalance

Rare variants are often geographically localized, making ancestry confounding especially important. Principal components or other population variables may be included, cohorts may be stratified and meta-analyzed, or mixed-model approaches may be selected. Related participants cannot simply be treated as independent observations because this may inflate test statistics. Extreme case-control imbalance and very small carrier counts also require methods and interpretation suited to sparse data.

Population Structure Analysis can support ancestry review, while Genome-wide Association Analysis provides complementary common-variant evidence. Rare variant analysis remains a distinct set-based workflow rather than an extra filter applied to a GWAS result.

Illustrative rare variant association results with quality control, gene-level association plot, quantile plot, mask comparison, and contributing variant summaries. Figure 4: Representative outputs connect gene-level signals to model calibration, masks, carrier counts, and contributing variants.

Deliverables

Sample and variant QC summaries with exclusions
Annotated rare-variant table with cohort and external frequencies
Gene, region, or pathway set definitions and mask membership
Burden, SKAT, SKAT-O, or other agreed result tables
Effect, p-value, multiple-testing, carrier-count, and convergence fields where applicable
Gene-level association, quantile, mask-comparison, and candidate-detail figures
Prioritized candidates with contributing variants and sensitivity results
Reproducible methods and analysis report

Best For and Not For

Best For	Not For
WES/WGS case-control or quantitative-trait cohorts, gene-based discovery, pharmaceutical target and preclinical biomarker research, and external VCF plus phenotype projects with sufficient cohort information	Underspecified phenotypes, unharmonized case/control sequencing, variant sets defined after inspecting association results, or cohorts too small to support the planned set tests

Case Study: Gene-Based Rare Variant Associations with Somatic Mutation Processes

The impact of rare germline variants on human somatic mutation processes

Journal: Nature Communications

Published: 2022

Vali-Pour M, Park S, Espinosa-Carrasco J, Ortiz-Martínez D, Lehner B, Supek F. The impact of rare germline variants on human somatic mutation processes. Nature Communications. 2022;13:3724.

The authors investigated whether inherited rare variants influence the rates and patterns of somatic mutation observed among cancer genomes.

They performed gene-based rare variant association analyses across human cancer sequencing cohorts, using prioritized rare putative loss-of-function variant sets, multiple inheritance models, and a combined burden and variance test. Discovery results were retested in an independent validation cohort.

The study reported 207 associations that replicated at an empirical false discovery rate of 1%, spanning 42 genes and 15 somatic mutational phenotypes. The results show how predefined masks, combined tests, multiple-testing control, and independent validation can be linked in a rare-variant workflow.

Figure 5: Original case-study summary of rare variant grouping, association testing, and independent validation reported by Vali-Pour et al. (2022).

The study illustrates a rigorous pattern for rare variant research: biologically defined variant sets, models that accommodate different genetic architectures, explicit multiple testing, and independent replication. Its findings are specific to the analyzed cohorts and phenotypes.

Why CD Genomics for Rare Variant Association Analysis

CD Genomics connects cohort QC, annotation, variant-set construction, statistical modeling, and candidate review so that a gene-level p-value can be traced back to the samples, variants, masks, and assumptions that produced it.

Predefined analysis specifications: frequency thresholds, annotation masks, grouping rules, tests, covariates, and correction procedures are documented before result review.

Architecture-aware testing: burden, SKAT, and SKAT-O are selected or compared based on plausible causal proportions and effect directions.

Cohort-aware calibration: ancestry, relatedness, batch, case-control imbalance, missingness, and sparse carrier counts are evaluated as part of the model.

Traceable candidates: results connect each gene or region to contributing variants, carrier counts, annotations, masks, and sensitivity analyses.

FAQs

What MAF threshold defines a rare variant?

There is no universal threshold. The cutoff should reflect sample size, ancestry, external frequency resources, expected genetic architecture, and method. One or more prespecified thresholds may be analyzed as separate masks.

How do you choose between burden, SKAT, and SKAT-O?

Burden tests favor aligned effects and a high causal proportion; SKAT tolerates mixed directions and neutral variants; SKAT-O adapts between them. Choice depends on biology, mask specificity, cohort size, and phenotype.

Can external VCF files be analyzed?

Yes, if genotype fields, reference assembly, calling workflow, sample metadata, phenotype, covariates, and QC provenance are sufficient. Recalling or harmonization may be recommended when cases and controls were processed differently.

How are ancestry and relatedness handled?

Options include ancestry covariates, stratified analysis and meta-analysis, mixed models, and relationship matrices. The appropriate approach depends on cohort structure and the selected association method.

How much sample size is needed?

Power depends on phenotype prevalence or variance, carrier counts, set size, causal proportion, effect magnitude and direction, ancestry, covariates, and testing burden. Feasibility should be evaluated through expected carriers and power simulations when possible.

Generate or process cohort data through Whole Exome Sequencing, Whole Genome Resequencing, and Variant Calling. Complement gene-based tests with Genome-wide Association Analysis and Population Structure Analysis.

References

Li B, Leal SM. Methods for detecting associations with rare variants for common diseases: application to analysis of sequence data. American Journal of Human Genetics. 2008;83(3):311-321. View article by DOI.
Wu MC, Lee S, Cai T, Li Y, Boehnke M, Lin X. Rare-variant association testing for sequencing data with the sequence kernel association test. American Journal of Human Genetics. 2011;89(1):82-93. View article by DOI.
Lee S, Emond MJ, Bamshad MJ, et al. Optimal unified approach for rare-variant association testing with application to small-sample case-control whole-exome sequencing studies. American Journal of Human Genetics. 2012;91(2):224-237. View article by DOI.
Chen H, Huffman JE, Brody JA, et al. Efficient variant set mixed model association tests for continuous and binary traits in large-scale whole-genome sequencing studies. American Journal of Human Genetics. 2019;104(2):260-274. View article by DOI.
Li X, Li Z, Zhou H, et al. Dynamic incorporation of multiple in silico functional annotations empowers rare variant association analysis of large whole-genome sequencing studies at scale. Nature Genetics. 2020;52:969-983. View article by DOI.
Vali-Pour M, Park S, Espinosa-Carrasco J, Ortiz-Martínez D, Lehner B, Supek F. The impact of rare germline variants on human somatic mutation processes. Nature Communications. 2022;13:3724. View article by DOI.