Motif analysis using topic modeling and DA results for Buenrostro et al (2018) scATAC-seq result

Last updated: 2022-03-29

Checks: 7 0

Knit directory: scATACseq-topics/

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(20200729)

The command set.seed(20200729) 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: b94b6bd

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 b94b6bd. 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:    .Rhistory
    Ignored:    .Rproj.user/

Untracked files:
    Untracked:  analysis/motif_analysis_Buenrostro2018_v2.Rmd
    Untracked:  output/clustering-Cusanovich2018.rds
    Untracked:  paper/
    Untracked:  scripts/DA_Buenrostro2018_filteredpeaks.sbatch
    Untracked:  scripts/postfit_Buenrostro2018_filteredpeaks.sbatch
    Untracked:  scripts/postfit_Buenrostro2018_v2.sbatch

Unstaged changes:
    Modified:   analysis/assess_fits_Buenrostro2018_Chen2019pipeline.Rmd
    Modified:   analysis/clusters_Cusanovich2018_k13.Rmd
    Modified:   analysis/gene_analysis_Buenrostro2018_Chen2019pipeline.Rmd
    Modified:   analysis/gene_analysis_Cusanovich2018.Rmd
    Modified:   analysis/index.Rmd
    Modified:   analysis/motif_analysis_Buenrostro2018_Chen2019pipeline.Rmd
    Modified:   analysis/motif_analysis_Cusanovich2018.Rmd
    Modified:   analysis/plots_Cusanovich2018.Rmd
    Modified:   scripts/DA_analysis_Buenrostro2018_vsnull.R
    Modified:   scripts/analysis_Buenrostro2018_filteredpeaks.sh

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/motif_analysis_Buenrostro2018_k10_filterpeaks.Rmd) and HTML (docs/motif_analysis_Buenrostro2018_k10_filterpeaks.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	b94b6bd	kevinlkx	2022-03-29	checked shared regions
html	2dbe21c	kevinlkx	2022-03-29	Build site.
Rmd	20907b9	kevinlkx	2022-03-29	added motif enrichment result using the top 1% regions
html	0ce1017	kevinlkx	2022-03-29	Build site.
Rmd	2b53fee	kevinlkx	2022-03-29	added motif enrichment result using the top 1% regions
html	85d9415	kevinlkx	2022-03-29	Build site.
Rmd	40f690f	kevinlkx	2022-03-29	updated DA and motif analysis using data with filtered peaks

Here we perform TF motif enrichment analysis for the Buenrostro et al (2018) scATAC-seq result inferred from the multinomial topic model with \(k = 10\).

We use binarized data downloaded from original paper.

Load packages and some functions used in this analysis

library(Matrix)
library(fastTopics)
library(dplyr)
library(tidyr)
library(ggplot2)
library(ggrepel)
library(cowplot)
library(plotly)
library(DT)
library(reshape2)
source("code/motif_analysis.R")
source("code/plots.R")

Load data and topic model results

Data downloaded from original paper.

data.dir <- "/project2/mstephens/kevinluo/scATACseq-topics/data/Buenrostro_2018/processed_data/"
load(file.path(data.dir, "Buenrostro_2018_binarized.RData"))
cat(sprintf("%d x %d counts matrix.\n",nrow(counts),ncol(counts)))
samples$cell <- rownames(samples)
samples$label <- as.factor(samples$label)

# 2034 x 465536 counts matrix.

Filtered out peaks with accessbility in fewer than 20 cells

i      <- which(colSums(counts) >= 20)
peaks  <- peaks[i,]
counts <- counts[,i]
cat(sprintf("After filtering, we have %d rows and %d columns left. \n",nrow(counts),ncol(counts)))

# After filtering, we have 2034 rows and 126719 columns left.

Load data after filtering peaks

res.dir <- "/project2/mstephens/kevinluo/scATACseq-topics/output/Buenrostro_2018/binarized/filtered_peaks"
load(file.path(res.dir, "Buenrostro_2018_binarized_filtered.RData"))
cat(sprintf("%d x %d counts matrix.\n",nrow(counts),ncol(counts)))
samples$cell <- rownames(samples)
samples$label <- as.factor(samples$label)

# 2034 x 126719 counts matrix.

Load the K = 10 topic model fit to data after peak filtering

fit <- readRDS(file.path(res.dir, "/fit-Buenrostro2018-binarized-filtered-scd-ex-k=10.rds"))$fit
fit <- poisson2multinom(fit)

Structure plot

topic_colors <- c("darkorange","limegreen","magenta","gold","skyblue",
                  "darkblue","dodgerblue","darkmagenta","red","olivedrab")

set.seed(1)
# labels <- factor(samples$label, levels = c("HSC", "MPP", "CMP", "GMP", "mono", "MEP", "LMPP", "CLP", "pDC", "UNK"))

labels <- factor(samples$label, c("mono","pDC","MEP","HSC","MPP","CLP",
                                 "LMPP","CMP","GMP","UNK"))
structure_plot(fit,grouping = labels,colors = topic_colors,
               # topics = 1:10,
               gap = 20,perplexity = 50,verbose = FALSE)

Version	Author	Date
85d9415	kevinlkx	2022-03-29

Differential accessibility (DA) analysis

We performed differenital accessbility (DA) analysis on topic modeling result with \(K = 10\)
DA analysis settings: ns = 10000, nsplit =100, shrink.method = "ash", lfc.stat = "vsnull", i.e. comparing each topic with the average.

Load DA analysis results with "ash" shrinkage (10000 MCMC iterations)

DA_dir <- "/project2/mstephens/kevinluo/scATACseq-topics/output/Buenrostro_2018/binarized/filtered_peaks/DAanalysis-Buenrostro2018-k=10"

DA_res <- readRDS(file.path(DA_dir, "DA_regions_topics_vsnull_ash_10000iters.rds"))
summary(DA_res)

dim(DA_res$z)

#          Length  Class  Mode   
# ar       1267190 -none- numeric
# est      1267190 -none- numeric
# postmean 1267190 -none- numeric
# lower    1267190 -none- numeric
# upper    1267190 -none- numeric
# z        1267190 -none- numeric
# lfsr     1267190 -none- numeric
# lpval          1 -none- numeric
# svalue   1267190 -none- numeric
# ash            3 ash    list   
# F        1267190 -none- numeric
# f0        126719 -none- numeric
# [1] 126719     10

Number of regions selected at different lfsr cutoffs:

n.sig.regions <- matrix(NA, nrow = 10, ncol = 3)
colnames(n.sig.regions) <- c("lfsr < 0.01", "lfsr < 0.05", "lfsr < 0.1")
rownames(n.sig.regions) <- paste("topic", 1:nrow(n.sig.regions))

for(k in 1:10){
  lfsr <- DA_res$lfsr[,k]
  n.sig.regions[k, ] <- c(length(which(lfsr < 0.01)), length(which(lfsr < 0.05)), length(which(lfsr < 0.1)))
}

n.sig.regions

#          lfsr < 0.01 lfsr < 0.05 lfsr < 0.1
# topic 1        15066       21034      25515
# topic 2        11164       18536      24600
# topic 3        15139       20590      25533
# topic 4        14210       19177      23482
# topic 5        15240       22228      26977
# topic 6         6892        9588      11650
# topic 7        13505       19528      24046
# topic 8         9983       17102      22741
# topic 9        13028       18154      22778
# topic 10        6605       10974      14385

Volcano plots for DA result

plots <- vector("list",10)
names(plots) <- 1:10

for (k in 1:10)
  plots[[k]] <- volcano_plot(DA_res, k, labels = rep("",nrow(DA_res$z)))
do.call(plot_grid,plots)

Version	Author	Date
85d9415	kevinlkx	2022-03-29

Motif enrichment analysis using HOMER

About HOMER motif analysis
- http://homer.ucsd.edu/homer/motif/
- http://homer.ucsd.edu/homer/ngs/peakMotifs.html
Script for motif enrichment analysis using HOMER
We performed differenital accessbility (DA) analysis with lfc.stat = "vsnull", i.e. comparing each topic with the average across topics.

Motif enrichment result using DA regions with lsfr < 0.01

We selected the significant peaks (regions) with lsfr < 0.01 to perform motif enrichment analysis using HOMER.

Load and compile HOMER results across topics

homer.dir <- "/project2/mstephens/kevinluo/scATACseq-topics/output/Buenrostro_2018/binarized/filtered_peaks/motifanalysis-Buenrostro2018-k=10-vsnull-ash/HOMER/DA_lfsr_0.01_regions"
cat(sprintf("Directory of motif analysis result: %s \n", homer.dir))
homer_res_topics <- readRDS(file.path(homer.dir, "/homer_knownResults.rds"))
selected_regions <- readRDS(file.path(homer.dir, "/selected_regions.rds"))

# Compile Homer results (pvalue and ranking) across topics
motif_res <- compile_homer_motif_res(homer_res_topics)
saveRDS(motif_res, paste0(homer.dir, "/homer_motif_enrichment_results.rds"))
cat("compiled homer motif results are saved in", paste0(homer.dir, "/homer_motif_enrichment_results.rds \n"))

# Directory of motif analysis result: /project2/mstephens/kevinluo/scATACseq-topics/output/Buenrostro_2018/binarized/filtered_peaks/motifanalysis-Buenrostro2018-k=10-vsnull-ash/HOMER/DA_lfsr_0.01_regions 
# compiled homer motif results are saved in /project2/mstephens/kevinluo/scATACseq-topics/output/Buenrostro_2018/binarized/filtered_peaks/motifanalysis-Buenrostro2018-k=10-vsnull-ash/HOMER/DA_lfsr_0.01_regions/homer_motif_enrichment_results.rds

List top 10 motifs for each topic

cat("Number of regions selected for each topic: \n")
print(mapply(nrow, selected_regions[1:(length(selected_regions)-1)]))

colnames_homer <- c("motif_name", "consensus", "P", "log10P", "Padj",  "num_target", "percent_target", "num_bg", "percent_bg")

top_motifs <- data.frame(matrix(nrow=10, ncol = length(homer_res_topics)))
colnames(top_motifs) <- names(homer_res_topics)
for (k in 1:length(homer_res_topics)){
  homer_res <- homer_res_topics[[k]]
  colnames(homer_res) <- colnames_homer
  homer_res <- homer_res %>% separate(motif_name, c("motif", "origin", "database"), "/")
  top_motifs[,k] <- head(homer_res$motif, 10)
}

DT::datatable(data.frame(rank = 1:10, top_motifs), rownames = F, caption = "Top 10 motifs enriched in each topic.")

# Number of regions selected for each topic: 
#    k1    k2    k3    k4    k5    k6    k7    k8    k9   k10 
# 15066 11164 15139 14210 15240  6892 13505  9983 13028  6605

Heatmap of motif enrichment -log10(p-value).

create_motif_enrichment_heatmap(motif_res, enrichment = "-log10(p-value)", 
                                cluster_motifs = TRUE, cluster_topics = TRUE, motif_filter = 10, horizontal = FALSE,
                                enrichment_range = c(0,100), method_cluster = "average", font.size.motifs = 4, font.size.topics = 9)

Version	Author	Date
0ce1017	kevinlkx	2022-03-29

# 220 out of 439 motifs included the heatmap

Rank the top enriched motifs for each topic

plots <- vector("list", ncol(motif_res$mlog10P))
names(plots) <- colnames(motif_res$mlog10P)

for( i in 1:length(plots)){
  plots[[i]] <- create_motif_enrichment_ranking_plot(motif_res, k = i, 
                                                     max.overlaps = 50, subsample = FALSE)
}

do.call(plot_grid,plots)

Version	Author	Date
2dbe21c	kevinlkx	2022-03-29
0ce1017	kevinlkx	2022-03-29

Motif enrichment result using top 1% DA regions with the largest logFC

We selected the top 1% peaks (regions) with the largest logFC to perform motif enrichment analysis using HOMER.