Last updated: 2022-03-29

Checks: 7 0

Knit directory: logistic-susie-gsea/

This reproducible R Markdown analysis was created with workflowr (version 1.7.0). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.

R Markdown file: up-to-date

Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.

Environment: empty

Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.

Seed: set.seed(20220105)

The command set.seed(20220105) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.

Session information: recorded

Great job! Recording the operating system, R version, and package versions is critical for reproducibility.

Cache: none

Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.

File paths: relative

Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.

Repository version: 122deec

Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.

The results in this page were generated with repository version 122deec. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.

Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated: 


Ignored files:
    Ignored:    .DS_Store
    Ignored:    .RData
    Ignored:    .Rhistory
    Ignored:    .Rproj.user/
    Ignored:    library/
    Ignored:    renv/library/
    Ignored:    renv/staging/
    Ignored:    staging/

Untracked files:
    Untracked:  _targets.R
    Untracked:  _targets.html
    Untracked:  _targets.md
    Untracked:  _targets/
    Untracked:  _targets_r/
    Untracked:  analysis/fetal_reference_cellid_gsea.Rmd
    Untracked:  analysis/fixed_intercept.Rmd
    Untracked:  analysis/iDEA_examples.Rmd
    Untracked:  analysis/latent_gene_list.Rmd
    Untracked:  analysis/latent_logistic_susie.Rmd
    Untracked:  analysis/libra_setup.Rmd
    Untracked:  analysis/linear_method_failure_modes.Rmd
    Untracked:  analysis/linear_regression_failure_regime.Rmd
    Untracked:  analysis/logistic_susie_veb_boost_vs_vb.Rmd
    Untracked:  analysis/references.bib
    Untracked:  analysis/simulations.Rmd
    Untracked:  analysis/test.Rmd
    Untracked:  analysis/wenhe_baboon_example.Rmd
    Untracked:  build_site.R
    Untracked:  cache/
    Untracked:  code/latent_logistic_susie.R
    Untracked:  code/marginal_sumstat_gsea_collapsed.R
    Untracked:  data/adipose_2yr_topsnp.txt
    Untracked:  data/fetal_reference_cellid_gene_sets.RData
    Untracked:  data/pbmc-purified/
    Untracked:  docs.zip
    Untracked:  index.md
    Untracked:  latent_logistic_susie_cache/
    Untracked:  simulation_targets/
    Untracked:  single_cell_pbmc_cache/
    Untracked:  summary_stat_gsea_exploration_cache/

Unstaged changes:
    Modified:   _simulation_targets.R
    Modified:   _targets.Rmd
    Modified:   analysis/gseabenchmark_tcga.Rmd
    Modified:   code/fit_baselines.R
    Modified:   code/fit_logistic_susie.R
    Modified:   code/fit_mr_ash.R
    Modified:   code/fit_susie.R
    Modified:   code/load_gene_sets.R
    Modified:   code/marginal_sumstat_gsea.R
    Modified:   code/simulate_gene_lists.R
    Modified:   code/utils.R
    Modified:   target_components/factories.R
    Modified:   target_components/methods.R

Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.

These are the previous versions of the repository in which changes were made to the R Markdown (analysis/single_cell_pbmc.Rmd) and HTML (docs/single_cell_pbmc.html) files. If you’ve configured a remote Git repository (see ?wflow_git_remote), click on the hyperlinks in the table below to view the files as they were in that past version.

File Version Author Date Message
Rmd 122deec karltayeb 2022-03-29 wflow_publish(pages)

Introduction

Our goals here are to run Logistic SuSiE on differential expression results from TCGA. We want to assess: 1. If the resulting enrichment results look good/interpretable across multiple/concatenated gene sets 2. Assess sensitivity to a range of p-value thresholds 3. Evaluate the potential of the summary stat latent model

library(GSEABenchmarkeR)
library(EnrichmentBrowser)
library(tidyverse)
library(susieR)
library(DT)
source('code/load_gene_sets.R')
source('code/utils.R')

source('code/logistic_susie_vb.R')
source('code/logistic_susie_veb_boost.R')
source('code/latent_logistic_susie.R')

Load Gene Sets

loadGeneSetX uniformly formats gene sets and generates the \(X\) matrix We can source any gene set from WebGestaltR::listGeneSet()

gs_list <- WebGestaltR::listGeneSet()
gobp <- loadGeneSetX('geneontology_Biological_Process', min.size=50)  # just huge number of gene sets
gobp_nr <- loadGeneSetX('geneontology_Biological_Process_noRedundant', min.size=1)
gomf <- loadGeneSetX('geneontology_Molecular_Function', min.size=1)
kegg <- loadGeneSetX('pathway_KEGG', min.size=1)
reactome <- loadGeneSetX('pathway_Reactome', min.size=1)
wikipathway_cancer <- loadGeneSetX('pathway_Wikipathway_cancer', min.size=1)
wikipathway <- loadGeneSetX('pathway_Wikipathway', min.size=1)

genesets <- list(
  gobp=gobp,
  gobp_nr=gobp_nr,
  gomf=gomf,
  kegg=kegg,
  reactome=reactome,
  wikipathway_cancer=wikipathway_cancer,
  wikipathway=wikipathway
)
load('data/pbmc-purified/deseq2-pbmc-purified.RData')

convert_labels <- function(y, from='SYMBOL', to='ENTREZID'){
  hs <- org.Hs.eg.db::org.Hs.eg.db
  gene_symbols <- names(y)
  symbol2entrez <- AnnotationDbi::select(hs, keys=gene_symbols, columns=c(to, from), keytype = from)
  symbol2entrez <- symbol2entrez[!duplicated(symbol2entrez[[from]]),]
  symbol2entrez <- symbol2entrez[!is.na(symbol2entrez[[to]]),]
  symbol2entrez <- symbol2entrez[!is.na(symbol2entrez[[from]]),]
  rownames(symbol2entrez) <- symbol2entrez[[from]]
  ysub <- y[names(y) %in% symbol2entrez[[from]]]
  names(ysub) <- symbol2entrez[names(ysub),][[to]]
  return(ysub)
}


par(mfrow=c(1,1))
deseq$`CD19+ B` %>% .$padj %>% hist(main='CD19+B p-values')
Loading required package: DESeq2

logistic_susie_driver = function(db, celltype, thresh){
  gs <- genesets[[db]]
  data <- deseq[[celltype]]
  
  # set up binary y
  y <- data %>%
    as.data.frame %>%
    rownames_to_column('gene') %>%
    dplyr::select(gene, padj) %>%
    filter(!is.na(padj)) %>%
    mutate(y = as.integer(padj < thresh)) %>%
    select(gene, y) %>%
    tibble2namedlist %>%
    convert_labels('ENSEMBL')
  
  u <- process_input(gs$X, y)  # subset to common genes
  vb.fit <- logistic.susie(  # fit model
    u$X, u$y, L=10, init.intercept = 0, verbose=1, maxit=100)

  # summarise results
  set.summary <- vb.fit$pip %>% 
    as_tibble(rownames='geneSet') %>%
    rename(pip=value) %>%
    mutate(
      top_component = apply(vb.fit$alpha, 2, which.max),
      active_set = top_component %in% vb.fit$sets$cs_index,
      top_component = paste0('L', top_component),
      cs = purrr::map(top_component, ~tryCatch(
        colnames(gs$X)[get(.x, vb.fit$sets$cs)], error = function(e) list())),
      in_cs = geneSet %in% cs,
      beta = colSums(vb.fit$mu * vb.fit$alpha),
      geneListSize = sum(u$y),
      geneSetSize = colSums(u$X),
      overlap = (u$y %*% u$X)[1,],
      nGenes = length(u$y),
      propSetInList = overlap / geneSetSize,
      oddsRatio = (overlap / (geneListSize - overlap)) / (
        (geneSetSize - overlap) / (nGenes - geneSetSize + overlap)),
    pValueHypergeometric = phyper(
      overlap-1, geneListSize, nGenes, geneSetSize, lower.tail= FALSE),
    db = db,
    celltype = celltype,
    thresh = thresh
    ) %>% left_join(gs$geneSet$geneSetDes)
  return(list(fit = vb.fit, set.summary=set.summary))
}

Fit logistic SuSiE

For each celltype, we fit logistic SuSiE using multiple gene set sources at various threshold of padj. Our goal is to assess 1. The quality of the gene set enrichments we get from each celltype - do reported gene set enrichments seem celltype specific/celltype relevant? - how much “interesting” marginal enrichment do we fail to capture in the multivariate model - how sensitive are we to the choice of pvalue threshold

celltypes <- names(deseq)
pthresh <- c(0.1, 0.01, 0.001, 0.0001, 0.00001, 0.000001)
db_name <- names(genesets)
crossed <- cross3(db_name, celltypes, pthresh)

pbmc_res <- xfun::cache_rds({
  res <- purrr::map(crossed, purrr::lift_dl(logistic_susie_driver))
  for (i in 1:length(res)){  # save some space
    res[[i]]$fit$dat <- NULL
  }
  res
  }, file = 'logistic_susie_pbmc_genesets_pthresh.rds'
)

pbmc_res_set_summary <- dplyr::bind_rows(purrr::map(pbmc_res, ~ pluck(.x, 'set.summary')))

Explore enrichments

Summary functions

pval_focussed_table = function(thresh=1e-3, filter_db=NULL, filter_celltype=NULL, top.n=50){
  pbmc_res_set_summary %>%
  filter(
    case_when(
      is.null(filter_db) ~ TRUE,
      !is.null(filter_db) ~ db %in% filter_db
    ) &
    thresh == thresh &
    case_when(
      is.null(filter_celltype) ~ TRUE,
      !is.null(filter_celltype) ~ celltype %in% filter_celltype
    )
  )  %>%
  dplyr::arrange(celltype, db, pValueHypergeometric) %>%
  group_by(celltype, db) %>% slice(1:top.n) %>%
  select(celltype, db, geneSet, description, pip, top_component, oddsRatio, propSetInList, pValueHypergeometric) %>%
  mutate_at(vars(celltype, db), factor) %>%
  datatable(filter = 'top')
}

set_focussed_table = function(thresh=1e-3, filter_db=NULL, filter_celltype=NULL){
  pbmc_res_set_summary %>%
  filter(
    case_when(
      is.null(filter_db) ~ TRUE,
      !is.null(filter_db) ~ db %in% filter_db
    ) &
    thresh == 1e-3 &
    in_cs & active_set &
    case_when(
      is.null(filter_celltype) ~ TRUE,
      !is.null(filter_celltype) ~ celltype %in% filter_celltype
    )
  )  %>%
  dplyr::arrange(celltype, db, desc(pip)) %>%
  select(celltype, db, geneSet, description, pip, top_component, oddsRatio, propSetInList, pValueHypergeometric) %>%
  mutate_at(vars(celltype, geneSet, db), factor) %>%
  datatable(filter = 'top')
}

Results

padj < 1e-2

set_focussed_table(1e-2)

padj < 1e-3

set_focussed_table(1e-3)

padj < 1e-4

set_focussed_table(1e-4)

padj < 1e-5

set_focussed_table(1e-5)
knitr::knit_exit()