In neurogenomics/MAGMA_Celltyping: Find Causal Cell-Types Underlying Complex Trait Genetics

library(MAGMA.Celltyping) 
library(dplyr)

Intro

MAGMA.Celltyping is a software package that facilitates conducting cell-type-specific enrichment tests on GWAS summary statistics.

Setup

Specify where you want the large files to be downloaded to.

NOTE: Make sure you change storage_dir to somewhere other than tempdir() if you want to make sure the results aren't deleted after this R session closes!

storage_dir <- tempdir()

Prepare data

GWAS

• We need to have a summary statistics file to analyse as input.
• As an example, you can download UK Biobank summary statistics for 'fluid_intelligence' using get_example_gwas().

Here we provide a pre-munged version of the above file.

Munging

Our lab have created MungeSumstats, a robust Bioconductor package for formatting multiple types of summary statistics files. We highly recommend processing your GWAS summary statistics with MungeSumstats before continuing. See the full_workflow vignette for more details.

The minimum info needed after munging is:
- "SNP", "CHR", and "BP" as first three columns. - It has at least one of these columns: "Z","OR","BETA","LOG_ODDS","SIGNED_SUMSTAT"

path_formatted <- MAGMA.Celltyping::get_example_gwas(
  trait = "prospective_memory")

Map SNPs to Genes

Note you can input the genome build of your summary statistics for this step or it can be inferred if left NULL:

genesOutPath <- MAGMA.Celltyping::map_snps_to_genes(
  path_formatted = path_formatted,
  genome_build = "GRCh37")

MAGMA_Files_Public

Rather than preprocessing the GWAS yourself, you can instead use the MAGMA_Files_Public database we have created. It contains pre-computed MAGMA SNP-to-genes mapping files for hundreds of GWAS.

You can browse which GWAS traits are available by looking at the provided metadata.csv file.

magma_dirs <- MAGMA.Celltyping::import_magma_files(ids = "ieu-a-298")

CellTypeDataset

ewceData provides a number of CellTypeDatasets (CTD) to be used a cell-type transcriptomic signature reference files.

If you want to create your own single-cell transcriptomic reference, you'll need to first convert it to CTD using the instructions found in the EWCE package documentation here.

ctd <- ewceData::ctd()

Note that the cell type dataset loaded in the code above is the Karolinksa cortex/hippocampus data only. For the full Karolinska dataset with hypothalamus and midbrain instead use the following:

ctd <- MAGMA.Celltyping::get_ctd("ctd_allKI")

Or for the DRONC seq or AIBS datasets use:

ctd <- get_ctd("ctd_Tasic")
ctd <- get_ctd("ctd_DivSeq")
ctd <- get_ctd("ctd_AIBS")
ctd <- get_ctd("ctd_DRONC_human")
ctd <- get_ctd("ctd_DRONC_mouse")
ctd <- get_ctd("ctd_BlueLake2018_FrontalCortexOnly")
ctd <- get_ctd("ctd_BlueLake2018_VisualCortexOnly")
ctd <- get_ctd("ctd_Saunders")

Run cell-type enrichment analyses

MAGMA.Celltyping offers a suite of functions for conducting various types of cell-type-specific enrichment tests on GWAS summary statistics.

The celltype_associations_pipeline wraps several functions that in previous versions of MAGMA.Celltyping had to be set up and run separately. These include:

• Linear enrichment: calculate_celltype_associations(EnrichmentMode = "linear") internally. Activated when run_linear=TRUE.
• Top 10% enrichment: Uses calculate_celltype_associations(EnrichmentMode = "Top 10%") internally. Activated when run_top10=TRUE.
• Conditional enrichment: Uses calculate_conditional_celltype_associations internally. Activated when run_conditional=TRUE.

Thus, celltype_associations_pipeline is designed to make these analyses easier to run.

MAGMA_results <- MAGMA.Celltyping::celltype_associations_pipeline(
  magma_dirs = magma_dirs,
  ctd = ctd,
  ctd_species = "mouse", 
  ctd_name = "Zeisel2015", 
  run_linear = TRUE, 
  run_top10 = TRUE)

We've also saved a pre-computed version of these results as a dataset:

MAGMA_results <- MAGMA.Celltyping::enrichment_results

Plot results

Merge results

merge_results imports each of the MAGMA enrichment results files and merges them into one so that they can easily be plotted and further analysed.

merged_results <- MAGMA.Celltyping::merge_results(
  MAGMA_results = MAGMA_results)
knitr::kable(merged_results)

Heatmap

Now we'll construct a heatmap visualizing the enrichment results, such that each GWAS is shown on the y-axis and each cell-type is shown on the x-axis. Results can be further facetted by what kind of test was run (linear, top10%, and/or conditional).

heat <- MAGMA.Celltyping::results_heatmap(
  merged_results = merged_results, 
  title = "Alzheimer's Disease (ieu-a-298) vs. nervous system cell-types (Zeisel2015)",
  fdr_thresh = 1)

Top results

Top phenotypes

Get the phenotypes with the greatest number of significant cell-type enrichment results.

top_phenos <- merged_results %>% 
  dplyr::group_by(EnrichmentMode, GWAS) %>%
  dplyr::summarise(Celltype=dplyr::n_distinct(Celltype)) %>%
  dplyr::arrange(dplyr::desc(Celltype))
knitr::kable(top_phenos)

Top enrichments

Get the phenotypes-celltype enrichment results with the most significant p-values (per phenotype).

top_enrich <- merged_results %>% 
  dplyr::group_by(EnrichmentMode, GWAS) %>%
  dplyr::slice_min(FDR, n = 2)
knitr::kable(top_enrich) 

Session Info

utils::sessionInfo()

---

neurogenomics/MAGMA_Celltyping documentation built on Oct. 12, 2024, 12:36 a.m.