BrainXcan explores the association between a complex phenotype and a collection of brain phenotypes quantified by MRI. Since these brain phenotypes are derived from image, we call them IDPs (image-derived phenotypes). In typical observation study, this is done by associating complex trait with observed brain IDPs.
But in BrainXcan, we consider using a study cohort from which we have both genotype and phenotype information (any GWAS cohort meets these requirements). First, we impute brain IDPs from the genotype (this is the so called genetically determined brain IDPs). And second, we associate the imputed brain IDPs with the phenotype.
It turns out that BrainXcan analysis can be performed using GWAS summary statistics without touching the individual-level data of the study cohort. BrainXcan software takes user-specified GWAS summary statistics as input and it performs:
Please refer to BrainXcan paper for more details.
BrainXcan software is implemented as a light-weight data pipeline written with snakemake. In below is shown the workflow.
sbxcan_t1
and sbxcan_dmri
run S-BrainXcan for all of the structural IDPs (T1 IDPs) and diffusion IDPs (dMRI IDPs) respectively.sbxcan_merge
aggregates the results from both structural and diffusion IDPs into one file.sbxcan_vis
visualizes the S-BrainXcan z-scores spatially.mr
runs MR for each significant IDPs (top 10 IDPs by default).mr_vis
visualizes and summarize the MR analysis for each significant IDP.sbxcan_report
summarizes all these results above into an automated report.In brief, a full S-BrainXcan run will contain the following outputs:
Please refer to “Get started with an example”.