Last updated: 2020-08-19
Checks: 7 0
Knit directory: scATACseq-topics/
This reproducible R Markdown analysis was created with workflowr (version 1.6.2). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.
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.
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.
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.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
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 23b953f. 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: .Rproj.user/
Untracked files:
Untracked: analysis/scATACseq_analysis_Lareau2019.Rmd
Untracked: analysis/scATACseq_data_Cusanovich2018.Rmd
Unstaged changes:
Modified: analysis/index.Rmd
Modified: analysis/references.Rmd
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/scATACseq_analysis_Cusanovich2018.Rmd) and HTML (docs/scATACseq_analysis_Cusanovich2018.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 | 23b953f | kevinlkx | 2020-08-19 | wflow_publish(“analysis/scATACseq_analysis_Cusanovich2018.Rmd”) |
Reference: Cusanovich, D., Hill, A., Aghamirzaie, D., Daza, R., Pliner, H., Berletch, J., Filippova, G., Huang, X., Christiansen, L., DeWitt, W., Lee, C., Regalado, S., Read, D., Steemers, F., Disteche, C., Trapnell, C., Shendure, J. (2018). A Single-Cell Atlas of In Vivo Mammalian Chromatin Accessibility Cell 174(5), 1 35. https://dx.doi.org/10.1016/j.cell.2018.06.052
Data were downloaded from the website: https://atlas.gs.washington.edu/mouse-atac/data/. They also provided detail descriptions about these datasets.
RCC directory: /project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/
The authors included nice tutorials on how to get started with analysis of sci-ATAC-seq data: http://atlas.gs.washington.edu/mouse-atac/docs/
R/python scripts referenced in this tutorial are available on: https://github.com/shendurelab/mouse-atac and downloaded to /project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/code_from_authors/mouse-atac on RCC.
Many of the initial steps of processing raw sci-ATAC-seq libraries used for this study are similar to their previous work of sci-ATAC-seq on Drosophila melanogaster embryos at 3 different stages of development. “Cusanovich, D., Reddington, J., Garfield, D. et al. The cis-regulatory dynamics of embryonic development at single-cell resolution. Nature 555, 538–542 (2018). https://doi.org/10.1038/nature25981”. Code and documentation on processing sequencing data can be found on the Fly ATAC Github. Documentation on various downstream steps can be found in the Fly ATAC documentation.
atac_matrix.binary.qc_filtered.rds (only QC filtered cells are included).library(Matrix)
## atac_matrix.binary.qc_filtered.rds: binarized peak by cell matrix in RDS format.
binarized_matrix <- readRDS("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/ATAC_matrices/atac_matrix.binary.qc_filtered.rds")
dim(binarized_matrix)[1] 436206 81173atac_matrix.tfidf.qc_filtered.peaks.txt.peaks_tfidf <- read.table("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/ATAC_matrices/atac_matrix.tfidf.qc_filtered.peaks.txt", header = FALSE, stringsAsFactors = TRUE)
cat(nrow(peaks_tfidf), "peaks used as input to TFIDF.")167013 peaks used as input to TFIDF.atac_matrix.tfidf.qc_filtered.rds. This dataset has rare peaks filtered out and is then normalized with TFIDF to allow for input to PCA/TSNE (only QC filtered cells are included).library(Matrix)
## atac_matrix.tfidf.qc_filtered.rds: TFIDF normalized peak by cell matrix in RDS format.
atac_matrix.tfidf <- readRDS("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/ATAC_matrices/atac_matrix.tfidf.qc_filtered.rds")
dim(atac_matrix.tfidf)[1] 167013 81173cell_metadata.txt: Metadata for cells in TSV format, including several features such as TSNE coordinates, cluster assignments, and cell type assignments. Columns in this data:
cell_metadata <- read.table("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/metadata/cell_metadata.txt", header = TRUE, stringsAsFactors = TRUE, sep = "\t")
dim(cell_metadata)[1] 81173 11cell_metadata[1:3,] cell tissue tissue.replicate cluster
1 AGCGATAGAACGAATTCGCCTCCGACGGCAGGACGT Lung Lung2_62216 10
2 AGCGATAGAACGAATTCGTTGGTAGTCGATAGAGGC Spleen Spleen_62016 10
3 AGCGATAGAACGCGCAGAAAGCTAGGTTAGGCGAAG BoneMarrow BoneMarrow_62016 10
subset_cluster tsne_1 tsne_2 subset_tsne1 subset_tsne2
1 1 16.25804 17.05674 -14.44873 -3.100071
2 1 15.28149 16.10859 -15.23808 1.123518
3 1 17.03914 14.69750 -16.88165 5.244815
id cell_label
1 clusters_10.cluster_1 Hematopoietic progenitors
2 clusters_10.cluster_1 Hematopoietic progenitors
3 clusters_10.cluster_1 Hematopoietic progenitorscell_metadata.tissue_freq_filtered.txt: Same as cell_metadata.txt, but removes cells in each tissue belonging to a cell_label that accounts for less than 0.5% of the cells in that tissue. These very low frequency labels are often not cell types expected their respective tissues and could be due to slight imperfections in clustering, for example. Provided matrices would need to be subsetted to match this set of cells if using this metadata.cell_metadata <- read.table("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/metadata/cell_metadata.tissue_freq_filtered.txt", header = TRUE, stringsAsFactors = TRUE, sep = "\t")
dim(cell_metadata)[1] 80254 11cell_metadata[1:3,] cell tissue tissue.replicate cluster
1 AGCGATAGAACGAATTCGCCTCCGACGGCAGGACGT Lung Lung2_62216 10
2 AGCGATAGAACGAATTCGTTGGTAGTCGATAGAGGC Spleen Spleen_62016 10
3 AGCGATAGAACGCGCAGAAAGCTAGGTTAGGCGAAG BoneMarrow BoneMarrow_62016 10
subset_cluster tsne_1 tsne_2 subset_tsne1 subset_tsne2
1 1 16.25804 17.05674 -14.44873 -3.100071
2 1 15.28149 16.10859 -15.23808 1.123518
3 1 17.03914 14.69750 -16.88165 5.244815
id cell_label
1 clusters_10.cluster_1 Hematopoietic progenitors
2 clusters_10.cluster_1 Hematopoietic progenitors
3 clusters_10.cluster_1 Hematopoietic progenitorscell_type_assignments.xlsx: Excel document with three tabs expected cell types, cell type markers, and Cell type assignments that contain a pairs of tissues and expected cell types, a list of positive markers for each cell type, and the table of cell type assignments with extra details about assignment criteria when applicable, respectively. This is meant to document justification for cell type assignments provided in cell_metadata.txt above.
peak_promoter_intersections.txt: Metadata for peaks-intersected TSS pairs in TSV format.
peak_promoter_intersections <- read.table("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/metadata/peak_promoter_intersections.txt", header = TRUE, stringsAsFactors = TRUE, sep = "\t")
dim(peak_promoter_intersections)[1] 62545 9peak_promoter_intersections[1:3,] peak_id peak_chr peak_start peak_end ensembl_id
1 chr1_3206330_3206546 chr1 3206330 3206546 ENSMUSG00000051951.5
2 chr1_3456171_3457239 chr1 3456171 3457239 ENSMUSG00000089699.1
3 chr1_3660494_3662750 chr1 3660494 3662750 ENSMUSG00000051951.5
gene_short_name ensembl_transcript_id biotype strand
1 Xkr4 ENSMUST00000162897.1 processed_transcript -
2 Gm1992 ENSMUST00000161581.1 antisense +
3 Xkr4 ENSMUST00000070533.4 protein_coding -They used the LSI (Latent Semantic Indexing) appraoch for dimensionality reduction: first, transform the data using the frequency-inverse document frequency transformation (TF-IDF), and then use singular value decomposition (SVD) on the TF-IDF matrix to generate a lower dimensional representation of the data. Introduction to LSI from wikipedia.
TF-IDF: first weight all the sites for individual cells by the total number of sites accessible in that cell (term frequency); then multiply these weighted values by log(1 + the inverse frequency of each site across all cells), the inverse document frequency. Introduction to TF-IDF from wikipedia.
This representation was then used as input for the T-SNE (t-distributed Stochastic Neighbor Embedding) analysis using Rtsne package in R.
The code below uses an example function atac_dim_reduction in dim_reduction.R from the authors’ Github page that given a matrix will do TFIDF, PCA, and t-SNE and return the resulting PCA and TSNE coordinates. Note that this function takes the binarized matrix and a site_frequency_threshold argument (default 0.03 or site observed in at least 3% of cells).
source('/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/code_from_authors/mouse-atac/dim_reduction/dim_reduction.R')
binarized_matrix <- readRDS("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/ATAC_matrices/atac_matrix.binary.qc_filtered.rds")
# This function outputs a list with two items pca_coords and tsne_coords, which contain the PCA and t-SNE coordinates as dataframes where the cell IDs are included as the rownames.
results.dim_reduction <- atac_dim_reduction(binarized_matrix, site_frequency_threshold=0.02)library(Matrix)
library(irlba)
dir_results <- "/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/results/"
dir.create(dir_results, showWarnings = FALSE, recursive = TRUE)
site_frequency_threshold <- 0.02
atac_matrix <- readRDS("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/ATAC_matrices/atac_matrix.binary.qc_filtered.rds")
## select sites/peaks observed in at least (site_frequency_threshold)% of cells
num_cells_ncounted <- Matrix::rowSums(atac_matrix)
threshold <- ncol(atac_matrix) * site_frequency_threshold
ncounts <- atac_matrix[num_cells_ncounted >= threshold,]
## Normalize the data with TF-IDF
nfreqs <- t(t(ncounts) / Matrix::colSums(ncounts)) # term frequency
tf_idf_counts <- nfreqs * log(1 + ncol(ncounts) / Matrix::rowSums(ncounts)) # term frequency * inverse document frequency
saveRDS(tf_idf_counts, paste0(dir_results, "/atac_matrix.tfidf.qc_filtered_freq", site_frequency_threshold, ".rds"))## LSI on ATAC-seq matrix
# atac_matrix: a peak x cell scATAC-seq matrix
# n_PCs: number of PCs (singular vectors) included
# site_frequency_threshold: site observed in at least % of cells
# adapted from the atac_dim_reduction function in https://github.com/shendurelab/mouse-atac/blob/master/dim_reduction/dim_reduction.R
LSI_atac <- function(atac_matrix, n_PCs = 50, site_frequency_threshold = 0.03){
library(Matrix)
library(irlba)
## select sites/peaks observed in at least (site_frequency_threshold)% of cells
num_cells_ncounted <- Matrix::rowSums(atac_matrix)
threshold <- ncol(atac_matrix) * site_frequency_threshold
ncounts <- atac_matrix[num_cells_ncounted >= threshold,]
## Normalize the data with TF-IDF
nfreqs <- t(t(ncounts) / Matrix::colSums(ncounts)) # term frequency
tf_idf_counts <- nfreqs * log(1 + ncol(ncounts) / Matrix::rowSums(ncounts)) # term frequency * inverse document frequency
## Do SVD
set.seed(0)
# use IRLBA algorithm to compute a partial SVD
# IRLBA: Fast Truncated Singular Value Decomposition and Principal Components Analysis for Large Dense and Sparse Matrices
SVD <- irlba(tf_idf_counts, n_PCs, n_PCs, maxit=1000)
d_diag <- matrix(0, nrow=length(SVD$d), ncol=length(SVD$d))
diag(d_diag) <- SVD$d
SVD_vd <- t(d_diag %*% t(SVD$v))
rownames(SVD_vd) <- colnames(atac_matrix)
colnames(SVD_vd) <- paste0('pca_', 1:ncol(SVD_vd))
# return SVD_vd: cell x PC matrix
return(SVD_vd)
}atac_LSI <- LSI_atac(atac_matrix, 50, 0.02)atac_matrix.binary.qc_filtered.rds (only QC filtered cells are included).library(Matrix)
## atac_matrix.binary.qc_filtered.rds: binarized peak by cell matrix in RDS format.
binarized_matrix <- readRDS("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/ATAC_matrices/atac_matrix.binary.qc_filtered.rds")
dim(binarized_matrix)[1] 436206 81173counts <- t(binarized_matrix)
cat(sprintf("Number of cells (samples): %d\n",nrow(counts)))Number of cells (samples): 81173cat(sprintf("Number of peaks: %d\n",ncol(counts)))Number of peaks: 436206cat(sprintf("Proportion of counts that are non-zero: %0.1f%%.\n",
100*mean(counts > 0)))Proportion of counts that are non-zero: 1.2%.# 100*nnzero(counts)/prod(dim(counts))cell_metadata.txt: Metadata for cells in TSV format, including several features such as TSNE coordinates, cluster assignments, and cell type assignments. Columns in this data:
cell_metadata <- read.table("/project2/mstephens/kevinluo/scATACseq-topics/data/Cusanovich_2018/metadata/cell_metadata.txt", header = TRUE, stringsAsFactors = TRUE, sep = "\t")
dim(cell_metadata)[1] 81173 11cat(sprintf("Number of cells (samples): %d\n",nrow(cell_metadata)))Number of cells (samples): 81173print(cell_metadata[1:3,]) cell tissue tissue.replicate cluster
1 AGCGATAGAACGAATTCGCCTCCGACGGCAGGACGT Lung Lung2_62216 10
2 AGCGATAGAACGAATTCGTTGGTAGTCGATAGAGGC Spleen Spleen_62016 10
3 AGCGATAGAACGCGCAGAAAGCTAGGTTAGGCGAAG BoneMarrow BoneMarrow_62016 10
subset_cluster tsne_1 tsne_2 subset_tsne1 subset_tsne2
1 1 16.25804 17.05674 -14.44873 -3.100071
2 1 15.28149 16.10859 -15.23808 1.123518
3 1 17.03914 14.69750 -16.88165 5.244815
id cell_label
1 clusters_10.cluster_1 Hematopoietic progenitors
2 clusters_10.cluster_1 Hematopoietic progenitors
3 clusters_10.cluster_1 Hematopoietic progenitors
sessionInfo()R version 3.5.1 (2018-07-02)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Scientific Linux 7.4 (Nitrogen)
Matrix products: default
BLAS/LAPACK: /software/openblas-0.2.19-el7-x86_64/lib/libopenblas_haswellp-r0.2.19.so
locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] Matrix_1.2-15 workflowr_1.6.2
loaded via a namespace (and not attached):
[1] Rcpp_1.0.4.6 lattice_0.20-38 rprojroot_1.3-2 digest_0.6.25
[5] later_1.0.0 grid_3.5.1 R6_2.4.1 backports_1.1.7
[9] git2r_0.27.1 magrittr_1.5 evaluate_0.14 stringi_1.4.6
[13] rlang_0.4.6 fs_1.3.1 promises_1.1.0 whisker_0.4
[17] rmarkdown_2.1 tools_3.5.1 stringr_1.4.0 glue_1.4.1
[21] httpuv_1.5.3.1 xfun_0.14 yaml_2.2.0 compiler_3.5.1
[25] htmltools_0.4.0 knitr_1.28