gsem

Rust implementation of GenomicSEM — multivariate LD Score Regression and Structural Equation Modeling on GWAS summary statistics.

Install

cargo install gsem

CLI

# Munge GWAS summary statistics
gsem munge --files trait1.txt trait2.txt --hm3 w_hm3.snplist --trait-names T1 T2

# Run multivariate LDSC
gsem ldsc --traits T1.sumstats.gz T2.sumstats.gz --ld eur_w_ld_chr/ --out ldsc.json

# Fit SEM model
gsem usermodel --covstruc ldsc.json --model "F1 =~ NA*V1 + V2; F1 ~~ 1*F1" --out sem.tsv

# Common factor
gsem commonfactor --covstruc ldsc.json --out cf.tsv

# Multivariate GWAS
gsem userGWAS --covstruc ldsc.json --sumstats merged.tsv --model "F1 =~ NA*V1 + V2; F1 ~ SNP"
gsem commonfactorGWAS --covstruc ldsc.json --sumstats merged.tsv

# Merge sumstats
gsem sumstats --files T1.sumstats.gz T2.sumstats.gz --ref eur_w_ld_chr/ --out merged.tsv

As a library

[dependencies]
gsem-ldsc = "0.1"
gsem-sem  = "0.1"

use gsem_ldsc::{ldsc, LdscConfig, TraitSumstats};
use gsem_sem::{syntax, model::Model, estimator};

Workspace crates

Citation

Grotzinger, A.D. et al. Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits. Nat Genet 51, 513–520 (2019).

License

GPL-3.0