Ancestry-Aware GWAS Pipeline

A comprehensive Nextflow pipeline for multi-ancestry genome-wide association studies (GWAS) with fine-mapping, meta-analysis, polygenic risk score calculation, and functional annotation. Optimized for admixed populations with Tractor, GENESIS, and SAIGE support.

Overview

This pipeline addresses the critical need for ancestry-aware genetic analysis by:

Ancestry-stratified GWAS: Runs separate GWAS for each major ancestry group (EUR, AFR, EAS, SAS, AMR, MID, AAC, AHI, LAT1, LAT2, HET)
Admixed-optimized analysis: GENESIS, SAIGE, and Tractor for local ancestry-aware GWAS
Multi-ancestry meta-analysis: Uses MR-MEGA to model ancestry-correlated heterogeneity
Within-ancestry fine-mapping: PolyFun+SuSIE for annotation-prioritized fine-mapping
Multi-ancestry fine-mapping: MG-FLASH-FM for related traits, SuSIE-ME for single traits
Ancestry-aware PRS: Multiple methods including PRS-CSx, GAUDI, and local ancestry-informed approaches
Heritability analysis: Ancestry-specific h² and cross-ancestry genetic correlations
Functional annotation: MAGMA, FUMA, LAVA, FLAMES with cell-type specificity
Diverse QTL colocalization: Curated datasets beyond GTEx for blood/immune, mQTL, caQTL, and more
Survival analysis: SPA-Cox for time-to-event GWAS

Key Features for Admixed Populations

Tractor - Local Ancestry-Aware GWAS

Tractor decomposes genetic effects by ancestral origin in admixed individuals:

African American (AAC): 2-way admixture (EUR, AFR)
Latino (LAT1+LAT2 combined): 3-way admixture (EUR, AFR, NAT)
- LAT1 = Mexican/Central American
- LAT2 = Caribbean/South American
- Combined for better statistical power

GENESIS - Admixed Population GWAS

R/Bioconductor package optimized for:

Complex population structure
Related individuals (with kinship matrix)
Heterogeneous ancestry

SAIGE

Optimized for:

Case-control imbalance
Rare variants
Large biobank data

Pipeline Architecture

┌─────────────────┐
│   Input Data    │
│  (Genotypes +   │
│   Phenotypes)   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐     ┌──────────────────────┐
│   QC + Ancestry │────▶│  Optional Ancestry   │
│    (GRAF-ANC)   │     │  Inference (or use   │
│                 │     │  pre-computed calls) │
└────────┬────────┘     └──────────────────────┘
         │
    ┌────┴────┬────────┬────────┬────────┬────────┬────────┬────────┐
    ▼         ▼        ▼        ▼        ▼        ▼        ▼        ▼
┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐
│  EUR  │ │  AFR  │ │  EAS  │ │  SAS  │ │  AMR  │ │  MID  │ │  AAC  │ │ LATINO│
│ GWAS  │ │ GWAS  │ │ GWAS  │ │ GWAS  │ │ GWAS  │ │ GWAS  │ │Tractor│ │Tractor│
└───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘
    │         │        │        │        │        │        │        │
    └────┬────┴────────┴────────┴────────┴────────┴────────┴────────┘
         │
         ▼
┌─────────────────┐
│   MR-MEGA       │
│  Meta-Analysis  │
└────────┬────────┘
         │
    ┌────┴─────────────┐
    ▼                  ▼
┌─────────────┐  ┌─────────────┐
│ Within-Anc  │  │ Multi-Anc   │
│ Fine-Map    │  │ Fine-Map    │
│ PolyFun+    │  │ MG-FLASH-FM │
│ SuSIE       │  │ or SuSIE-ME │
└──────┬──────┘  └──────┬──────┘
       │                │
       └────────┬───────┘
                ▼
┌───────────────────────────────────────────────────────────┐
│                    Downstream Analysis                     │
├─────────────┬─────────────┬──────────────┬────────────────┤
│     PRS     │ Heritability│ Colocalization│   Functional   │
│  PRS-CSx    │    LDSC     │ Diverse QTL  │     MAGMA      │
│  PRS-CS     │    GCTA     │   Datasets   │     FUMA       │
│   GAUDI     │  Cross-Anc  │              │     LAVA       │
│  LDpred2    │     rg      │              │    FLAMES      │
└─────────────┴─────────────┴──────────────┴────────────────┘

Quick Start

# Basic run with REGENIE
nextflow run main.nf \
  --input samplesheet.csv \
  --phenotype_file phenotypes.txt \
  --phenotype_cols "trait1,trait2" \
  --outdir results \
  -profile docker

# With pre-computed ancestry calls (recommended)
nextflow run main.nf \
  --input samplesheet.csv \
  --phenotype_file phenotypes.txt \
  --phenotype_cols "WBC,Lymphocytes,Monocytes" \
  --covariate_cols "age,sex,PC1,PC2,PC3,PC4,PC5" \
  --run_ancestry_inference false \
  --ancestry_calls_file ancestry_calls.tsv \
  --gwas_tool genesis \
  --kinship_matrix grm.rds \
  --outdir results \
  -profile singularity

# For admixed populations with Tractor
nextflow run main.nf \
  --input samplesheet.csv \
  --phenotype_file phenotypes.txt \
  --phenotype_cols "trait1" \
  --run_tractor true \
  --local_ancestry_files local_ancestry.msp \
  --tractor_aac_pops "EUR,AFR" \
  --tractor_lat_pops "EUR,AFR,NAT" \
  --outdir results \
  -profile singularity

# For survival/time-to-event analysis
nextflow run main.nf \
  --input samplesheet.csv \
  --phenotype_file phenotypes.txt \
  --phenotype_cols "leukemia" \
  --survival_analysis true \
  --time_col "time_to_event" \
  --event_col "event_status" \
  --outdir results \
  -profile singularity

Input Requirements

Samplesheet

Create a CSV file with genotype file paths:

sample_id,cohort,bed,bim,fam
study1,cohort_A,/path/to/study1.bed,/path/to/study1.bim,/path/to/study1.fam
study2,cohort_B,/path/to/study2.bed,/path/to/study2.bim,/path/to/study2.fam

Supported formats: PLINK1 (bed/bim/fam), PLINK2 (pgen/pvar/psam), BGEN, VCF

Phenotype File

Tab-separated file with sample IDs and phenotype values:

IID	trait1	trait2	age	sex	PC1	PC2	PC3	PC4	PC5
sample1	25.3	170.2	45	1	0.01	-0.02	0.005	0.001	-0.003
sample2	28.1	165.8	38	2	-0.01	0.03	-0.002	0.004	0.001

Pre-computed Ancestry Calls (Optional but Recommended)

If you have pre-computed ancestry calls, provide a TSV file:

IID	ancestry
sample1	EUR
sample2	AFR
sample3	AAC
sample4	LAT1

Key Parameters

Ancestry Options

Parameter	Default	Description
`--run_ancestry_inference`	`false`	Run ancestry inference in pipeline
`--ancestry_calls_file`	`null`	Pre-computed ancestry calls file
`--ancestry_method`	`graf-anc`	Method: graf-anc, admixture, pca
`--ancestry_groups`	`EUR,AFR,EAS,SAS,AMR,MID,AAC,AHI,LAT1,LAT2,HET`	Ancestry groups
`--min_ancestry_n`	`100`	Minimum samples per ancestry

GWAS Tools

Parameter	Default	Description
`--gwas_tool`	`regenie`	Tool: regenie, saige, bolt-lmm, plink2, genesis
`--gwas_model`	`additive`	Model: additive, dominant, recessive
`--kinship_matrix`	`null`	Pre-computed kinship matrix (for GENESIS)
`--p_threshold`	`5e-8`	Genome-wide significance threshold

Tractor (Local Ancestry-Aware GWAS)

Parameter	Default	Description
`--run_tractor`	`false`	Run Tractor for admixed populations
`--tractor_groups`	`AAC,LATINO`	Groups for Tractor (LATINO = LAT1+LAT2)
`--tractor_aac_pops`	`EUR,AFR`	Ancestral populations for African American
`--tractor_lat_pops`	`EUR,AFR,NAT`	Ancestral populations for Latino
`--local_ancestry_files`	`null`	Local ancestry MSP files (RFMix format)

Survival Analysis (SPA-Cox)

Parameter	Default	Description
`--survival_analysis`	`false`	Run time-to-event GWAS
`--time_col`	`null`	Time-to-event column name
`--event_col`	`null`	Event indicator column (1=event, 0=censored)

Meta-Analysis

Parameter	Default	Description
`--meta_method`	`mr-mega`	Method: mr-mega, metal, metasoft
`--mr_mega_env`	`false`	Include environmental variable

Fine-Mapping

Parameter	Default	Description
`--within_ancestry_fm`	`polyfun-susie`	Within-ancestry method
`--multi_ancestry_fm`	`mg-flash-fm`	Multi-ancestry method for related traits
`--single_trait_fm`	`susie-me`	Multi-ancestry method for single traits
`--related_traits`	`true`	Whether traits are related

Colocalization & QTL Datasets

Parameter	Default	Description
`--use_curated_qtl`	`true`	Use curated diverse QTL datasets
`--qtl_ancestry_matching`	`true`	Match QTL datasets to GWAS ancestry
`--custom_qtl_datasets`	`null`	Path to custom QTL samplesheet

Curated QTL datasets include:

Blood/Immune: OneK1K (14 immune cell types), DICE, BLUEPRINT, eQTLGen
African ancestry: GENOA (mQTL/eQTL), AFGR (caQTL)
Multi-ethnic: MESA (EUR, AFR, HISP, EAS), TOPMed
Methylation: GENOA mQTL, GoDMC, BLUEPRINT mQTL
Chromatin accessibility: DICE caQTL, BLUEPRINT hQTL
Single-cell: OneK1K, HCA Immune Atlas

PRS Methods

Parameter	Default	Description
`--prs_methods`	`prs-csx,prs-cs,gaudi,ldpred2`	PRS methods to run
`--prs_validation`	`true`	Validate PRS in target cohort
`--prs_best_method`	`true`	Determine best method per ancestry

Ancestry Groups

Code	Description	Tractor Support
EUR	European	-
AFR	African	-
EAS	East Asian	-
SAS	South Asian	-
AMR	Native American	-
MID	Middle Eastern	-
AAC	African American (admixed)	Yes (2-way: EUR,AFR)
AHI	American Hispanic (admixed)	Yes (3-way: EUR,AFR,NAT)
LAT1	Latino Type 1 (Mexican/Central American)	Combined as LATINO
LAT2	Latino Type 2 (Caribbean/South American)	Combined as LATINO
HET	Heterogeneous/Other	-

Output Structure

results/
├── ancestry/
│   ├── graf/                    # GRAF ancestry calls
│   └── stratified/              # Ancestry-stratified genotypes
├── gwas/
│   ├── {ancestry}/{trait}/      # Per-ancestry GWAS results
│   └── tractor/                 # Tractor results (AAC, LATINO)
│       ├── AAC/                 # 2-way admixture results
│       └── LATINO/              # 3-way admixture results (LAT1+LAT2)
├── survival/
│   └── {ancestry}/{trait}/      # SPA-Cox survival GWAS
├── meta_analysis/
│   └── {trait}/                 # MR-MEGA meta-analysis
├── fine_mapping/
│   ├── {ancestry}/{trait}/      # Within-ancestry fine-mapping
│   └── multi_ancestry/{trait}/  # Multi-ancestry fine-mapping
├── prs/
│   └── {trait}/                 # PRS weights and scores
├── heritability/
│   └── {ancestry}/{trait}/      # h² estimates
├── functional/
│   └── {trait}/                 # Functional annotations
├── colocalization/
│   └── {trait}/                 # Colocalization results
├── plots/
│   └── {trait}/                 # Manhattan, QQ, regional plots
└── reports/
    ├── multiqc_report.html      # MultiQC report
    └── ancestry_gwas_report.html # Custom pipeline report

Methods Description

Tractor

Local ancestry-aware GWAS that decomposes genetic effects by ancestral origin:

Extracts ancestry-specific haplotype dosages from local ancestry calls
Tests for ancestry-specific effects and heterogeneity
Particularly powerful for identifying ancestry-specific associations

GENESIS

R/Bioconductor package for:

Mixed model association testing accounting for population structure and relatedness
Score, Wald, and BinomiRare tests
PC-AiR for ancestry inference in related samples
PC-Relate for kinship estimation in structured populations

SPA-Cox

Saddlepoint approximation for Cox proportional hazards GWAS:

Efficient for large-scale survival analysis
Handles time-to-event outcomes (e.g., disease onset, death)
Accounts for censoring

MR-MEGA Meta-Analysis

Meta-regression of genetic association data:

Models allelic effects as a function of axes of genetic variation
Accounts for ancestry-correlated heterogeneity
Detects residual heterogeneity beyond ancestry

PolyFun+SuSIE Fine-Mapping

Annotation-informed fine-mapping:

PolyFun: Functional annotation-informed prior probabilities
SuSIE: Sum of Single Effects regression for multiple causal variants
95% credible set construction

MG-FLASH-FM & SuSIE-ME

Multi-ancestry fine-mapping:

MG-FLASH-FM: For related traits (shares information across traits)
SuSIE-ME: For single/unrelated traits

PRS-CSx & GAUDI

Cross-population PRS:

PRS-CSx: Coupled continuous shrinkage priors across ancestries
GAUDI: Local ancestry-informed PRS for admixed individuals

Citation

If you use this pipeline, please cite:

[Pipeline citation - to be added]

Key methods:
- REGENIE: Mbatchou et al. (2021) Nature Genetics
- SAIGE: Zhou et al. (2018) Nature Genetics
- GENESIS: Gogarten et al. (2019) Bioinformatics
- Tractor: Atkinson et al. (2021) Nature Genetics
- MR-MEGA: Magi et al. (2017) Nature Genetics
- PolyFun: Weissbrod et al. (2020) Nature Genetics
- SuSIE: Wang et al. (2020) J R Stat Soc
- PRS-CSx: Ruan et al. (2022) Nature Genetics
- GAUDI: Marnetto et al. (2020) Nature Communications

Requirements

Nextflow >= 23.04.0
Container runtime: Docker, Singularity, or Apptainer
OR Conda/Mamba for local execution

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please read our contributing guidelines first.

Name		Name	Last commit message	Last commit date
Latest commit History 13 Commits
assets		assets
bin		bin
conf		conf
examples		examples
modules/local		modules/local
subworkflows		subworkflows
README.md		README.md
main.nf		main.nf
nextflow.config		nextflow.config

Atkinson-Lab/Nextflow_GWAS_pipeline

Folders and files

Latest commit

History

Repository files navigation