Author: Xiurui Zhu
Modified: 2021-10-28 09:57:36
Compiled: 2021-10-28 09:57:42
Next-generation sequencing (NGS) is a powerful tool for analyzing gene sequences. There are a couple of packages that help to process NGS data, including sequence alignment data, gene annotation, experimental metadata and significance analysis. In this file, we will go through this process according to a tutorial for RNA-seq in R.
To facilitate the analyses in the workflow, we need to load the
following packages: tidyverse, magrittr, rlang, Rsamtools,
GenomicFeatures, GenomicAlignments, DESeq, DESeq2, GOstats,
GO.db, Category, org.At.tair.db and ComplexHeatmap.
# Define a function to check, install (if necessary) and load packages
check_packages <- function(pkg_name,
repo = c("cran", "github", "Bioconductor"),
repo_path) {
repo <- match.arg(repo)
# Load installed packages
inst_packages <- installed.packages()
if (pkg_name %in% inst_packages == FALSE) {
cat("* Installing: ", pkg_name, ", repo = ", repo, "\n", sep = "")
switch(repo,
cran = install.packages(pkg_name),
github = {
if ("devtools" %in% inst_packages == FALSE) {
install.packages("devtools")
}
devtools::install_github(repo_path)
},
Bioconductor = {
if ("BiocManager" %in% inst_packages == FALSE) {
install.packages("BiocManager")
}
BiocManager::install(pkg_name)
})
} else {
cat("* Package already installed: ", pkg_name, "\n", sep = "")
}
suppressPackageStartupMessages(
library(pkg_name, character.only = TRUE)
)
}
# CRAN packages
check_packages("tidyverse", repo = "cran")
purrr::walk(.x = c("magrittr", "rlang"),
.f = check_packages, repo = "cran")
purrr::walk(.x = c("Rsamtools", "GenomicFeatures",
"GenomicAlignments", "DESeq",
"DESeq2", "GOstats", "GO.db",
"Category", "org.At.tair.db",
"ComplexHeatmap"),
.f = check_packages, repo = "Bioconductor")
#> * Package already installed: tidyverse
#> * Package already installed: magrittr
#> * Package already installed: rlang
#> * Package already installed: Rsamtools
#> * Package already installed: GenomicFeatures
#> * Package already installed: GenomicAlignments
#> * Package already installed: DESeq
#> * Package already installed: DESeq2
#> * Package already installed: GOstats
#> * Package already installed: GO.db
#> * Package already installed: Category
#> * Package already installed: org.At.tair.db
#> * Package already installed: ComplexHeatmapAlignment data from 4 “bam” files were loaded as pointers to data on disk.
bam_files <- list.files("data", pattern = "\\.bam$", full.names = TRUE) %>%
Rsamtools::BamFileList()
print(bam_files)
#> BamFileList of length 4
#> names(4): KCL1_dedup_sorted.bam KCL2_dedup_sorted.bam NO31_dedup_sorted.bam NO32_dedup_sorted.bamThen, a “gtf” annotation file was loaded to locate genomic features and exons were grouped by genes.
tx_db <- list.files("data", pattern = "\\.gtf$", full.names = TRUE) %>%
GenomicFeatures::makeTxDbFromGFF(format="gtf")
#> Import genomic features from the file as a GRanges object ... OK
#> Prepare the 'metadata' data frame ... OK
#> Make the TxDb object ... OK
print(tx_db)
#> TxDb object:
#> # Db type: TxDb
#> # Supporting package: GenomicFeatures
#> # Data source: data/Arabidopsis.gtf
#> # Organism: NA
#> # Taxonomy ID: NA
#> # miRBase build ID: NA
#> # Genome: NA
#> # Nb of transcripts: 41671
#> # Db created by: GenomicFeatures package from Bioconductor
#> # Creation time: 2021-10-28 09:58:58 +0800 (Thu, 28 Oct 2021)
#> # GenomicFeatures version at creation time: 1.40.1
#> # RSQLite version at creation time: 2.2.8
#> # DBSCHEMAVERSION: 1.2
exon_by_gene <- tx_db %>%
GenomicFeatures::exonsBy(by = "gene")
print(exon_by_gene)
#> GRangesList object of length 33610:
#> $AT1G01010
#> GRanges object with 6 ranges and 2 metadata columns:
#> seqnames ranges strand | exon_id exon_name
#> <Rle> <IRanges> <Rle> | <integer> <character>
#> [1] Chr1 3631-3913 + | 1 <NA>
#> [2] Chr1 3996-4276 + | 2 <NA>
#> [3] Chr1 4486-4605 + | 3 <NA>
#> [4] Chr1 4706-5095 + | 4 <NA>
#> [5] Chr1 5174-5326 + | 5 <NA>
#> [6] Chr1 5439-5899 + | 6 <NA>
#> -------
#> seqinfo: 7 sequences (2 circular) from an unspecified genome; no seqlengths
#>
#> $AT1G01020
#> GRanges object with 12 ranges and 2 metadata columns:
#> seqnames ranges strand | exon_id exon_name
#> <Rle> <IRanges> <Rle> | <integer> <character>
#> [1] Chr1 5928-6263 - | 21880 <NA>
#> [2] Chr1 6437-7069 - | 21881 <NA>
#> [3] Chr1 6790-7069 - | 21882 <NA>
#> [4] Chr1 7157-7232 - | 21883 <NA>
#> [5] Chr1 7157-7450 - | 21884 <NA>
#> ... ... ... ... . ... ...
#> [8] Chr1 7762-7835 - | 21887 <NA>
#> [9] Chr1 7942-7987 - | 21888 <NA>
#> [10] Chr1 8236-8325 - | 21889 <NA>
#> [11] Chr1 8417-8464 - | 21890 <NA>
#> [12] Chr1 8571-8737 - | 21891 <NA>
#> -------
#> seqinfo: 7 sequences (2 circular) from an unspecified genome; no seqlengths
#>
#> $AT1G01030
#> GRanges object with 2 ranges and 2 metadata columns:
#> seqnames ranges strand | exon_id exon_name
#> <Rle> <IRanges> <Rle> | <integer> <character>
#> [1] Chr1 11649-13173 - | 21892 <NA>
#> [2] Chr1 13335-13714 - | 21893 <NA>
#> -------
#> seqinfo: 7 sequences (2 circular) from an unspecified genome; no seqlengths
#>
#> ...
#> <33607 more elements>Finally, experimental metadata were loaded for a description of the dataset.
exp_design <- file.path("data", "expdesign.txt") %>%
read.csv(row.names = 1L, sep = ",")
print(exp_design)
#> condition
#> KCL1_dedup_sorted.bam untreated
#> KCL2_dedup_sorted.bam untreated
#> NO31_dedup_sorted.bam treated
#> NO32_dedup_sorted.bam treatedReads were first counted with exons grouped by genes.
summ_exp <- GenomicAlignments::summarizeOverlaps(
features = exon_by_gene,
reads = bam_files,
mode = "Union",
singleEnd = TRUE,
ignore.strand = TRUE
)
print(summ_exp)
#> class: RangedSummarizedExperiment
#> dim: 33610 4
#> metadata(0):
#> assays(1): counts
#> rownames(33610): AT1G01010 AT1G01020 ... ATMG01400 ATMG01410
#> rowData names(0):
#> colnames(4): KCL1_dedup_sorted.bam KCL2_dedup_sorted.bam
#> NO31_dedup_sorted.bam NO32_dedup_sorted.bam
#> colData names(0):
sam_count <- SummarizedExperiment::assay(summ_exp)
dim(sam_count)
#> [1] 33610 4Median counts per condition in experimental metadata were counted and maximal median counts < 10 were removed.
count_thresh <- 10L
med_count <- sam_count %>%
as.data.frame() %>%
tibble::rownames_to_column("Feature_Name") %>%
tidyr::pivot_longer(cols = !c("Feature_Name"),
names_to = "File_Name",
values_to = "value") %>%
dplyr::inner_join(
exp_design %>%
tibble::rownames_to_column("File_Name"),
by = "File_Name"
) %>%
dplyr::group_by(Feature_Name, condition) %>%
dplyr::summarize(med_value = median(value, na.rm = TRUE),
.groups = "drop") %>%
tidyr::pivot_wider(id_cols = "Feature_Name",
names_from = "condition",
values_from = "med_value") %>%
tibble::column_to_rownames("Feature_Name") %>%
as.matrix()
max_med_count <- med_count %>%
as.data.frame() %>%
tibble::rownames_to_column("Feature_Name") %>%
dplyr::mutate(max_count = pmax(!!!dplyr::syms(colnames(med_count)),
na.rm = TRUE)) %>%
dplyr::select(Feature_Name, max_count) %>%
tibble::deframe()
sam_count_filter <- sam_count[
names(max_med_count)[max_med_count >= count_thresh],
1:ncol(sam_count)
]
dim(sam_count_filter)
#> [1] 11422 4Statistical analyses were carried out with DESeq package.
A count dataset (data container) was first constructed to analyze data
with DESeq. Then, size factors and variations were estimated.
count_deseq <- DESeq::newCountDataSet(
countData = sam_count_filter,
conditions = exp_design %>%
tibble::rownames_to_column("File_Name") %>%
tibble::deframe() %>%
as.factor() %>%
`[`(colnames(sam_count_filter))
) %>%
DESeq::estimateSizeFactors() %>%
DESeq::estimateDispersions()
print(count_deseq)
#> CountDataSet (storageMode: environment)
#> assayData: 11422 features, 4 samples
#> element names: counts
#> protocolData: none
#> phenoData
#> sampleNames: KCL1_dedup_sorted.bam KCL2_dedup_sorted.bam
#> NO31_dedup_sorted.bam NO32_dedup_sorted.bam
#> varLabels: sizeFactor condition
#> varMetadata: labelDescription
#> featureData
#> featureNames: AT1G01010 AT1G01040 ... AT5G67630 (11422 total)
#> fvarLabels: disp_pooled
#> fvarMetadata: labelDescription
#> experimentData: use 'experimentData(object)'
#> Annotation:
DESeq::plotDispEsts(count_deseq)
Next, univariate analysis was performed between the two groups define in
exp_design and a total of 208 significant genes were discovered.
count_univar <- count_deseq %>%
DESeq::nbinomTest("untreated", "treated")
head(count_univar)
#> id baseMean baseMeanA baseMeanB foldChange log2FoldChange pval
#> 1 AT1G01010 21.45700 27.32565 15.58834 0.5704655 -0.8097885 0.1238439
#> 2 AT1G01040 44.92871 39.63235 50.22508 1.2672748 0.3417294 0.3625881
#> 3 AT1G01050 14.11043 12.85400 15.36686 1.1954923 0.2576048 0.7524904
#> 4 AT1G01060 63.99045 71.62446 56.35645 0.7868324 -0.3458717 0.4788409
#> 5 AT1G01090 92.24328 80.91875 103.56782 1.2798990 0.3560299 0.2472544
#> 6 AT1G01100 180.46426 153.25272 207.67579 1.3551197 0.4384203 0.1022170
#> padj
#> 1 1
#> 2 1
#> 3 1
#> 4 1
#> 5 1
#> 6 1
sig_count <- sum(count_univar[["padj"]] < 0.05, na.rm=T)
print(sig_count)
#> [1] 208MA plot was used to indicate from the univariate results the required fold change for a gene to be significant against its average count.
DESeq::plotMA(count_univar, ylim = c(-5, 5))
Significant genes were derived from the univariate results.
sig_gene <- c(up = 1, down = -1) %>%
purrr::map(~ {
count_univar[count_univar[["padj"]] < 0.1 &
abs(count_univar[["log2FoldChange"]]) > 1 &
sign(count_univar[["log2FoldChange"]]) == .x, ]
})
tibble::glimpse(sig_gene)
#> List of 2
#> $ up :'data.frame': 204 obs. of 8 variables:
#> ..$ id : chr [1:204] "AT1G03850" "AT1G07150" "AT1G08090" "AT1G08100" ...
#> ..$ baseMean : num [1:204] 33.3 28 400.4 130.5 44.5 ...
#> ..$ baseMeanA : num [1:204] 15.09 4.04 81.96 24.05 19.16 ...
#> ..$ baseMeanB : num [1:204] 51.5 51.9 718.8 237 69.8 ...
#> ..$ foldChange : num [1:204] 3.41 12.83 8.77 9.85 3.64 ...
#> ..$ log2FoldChange: num [1:204] 1.77 3.68 3.13 3.3 1.87 ...
#> ..$ pval : num [1:204] 5.30e-05 3.34e-04 3.94e-06 4.78e-09 6.30e-06 ...
#> ..$ padj : num [1:204] 5.06e-03 2.33e-02 6.25e-04 1.48e-06 8.89e-04 ...
#> $ down:'data.frame': 33 obs. of 8 variables:
#> ..$ id : chr [1:33] "AT1G13110" "AT1G27030" "AT1G29050" "AT1G55760" ...
#> ..$ baseMean : num [1:33] 54.8 56.7 79.5 19 61.7 ...
#> ..$ baseMeanA : num [1:33] 76.6 79.8 109.6 33.2 110.3 ...
#> ..$ baseMeanB : num [1:33] 33.07 33.63 49.45 4.79 13.14 ...
#> ..$ foldChange : num [1:33] 0.432 0.421 0.451 0.144 0.119 ...
#> ..$ log2FoldChange: num [1:33] -1.21 -1.25 -1.15 -2.8 -3.07 ...
#> ..$ pval : num [1:33] 1.09e-03 1.02e-03 6.33e-04 2.25e-05 1.66e-04 ...
#> ..$ padj : num [1:33] 0.05763 0.05464 0.03745 0.00259 0.01306 ...
sig_gene_comb <- dplyr::bind_rows(sig_gene)
head(sig_gene_comb)
#> id baseMean baseMeanA baseMeanB foldChange log2FoldChange
#> 1 AT1G03850 33.27369 15.093992 51.45338 3.408865 1.769291
#> 2 AT1G07150 27.96515 4.044117 51.88618 12.830039 3.681454
#> 3 AT1G08090 400.35951 81.960358 718.75867 8.769589 3.132509
#> 4 AT1G08100 130.50403 24.051723 236.95633 9.851948 3.300409
#> 5 AT1G08650 44.49139 19.162379 69.82040 3.643618 1.865372
#> 6 AT1G11655 8.48047 1.704820 15.25612 8.948818 3.161697
#> pval padj
#> 1 5.300576e-05 5.058579e-03
#> 2 3.339807e-04 2.334359e-02
#> 3 3.937001e-06 6.245615e-04
#> 4 4.781152e-09 1.475954e-06
#> 5 6.301214e-06 8.885490e-04
#> 6 6.813422e-04 3.930450e-02The functions of genes were annotated with data from an open-source “tair” repository. Transcript identifiers ending with a dot and a number were removed for matching.
gene_desc <- list.files("data",
pattern = "^gene_description",
full.names = TRUE) %>%
readr::read_delim(col_types = readr::cols(),
delim = "\t",
col_names = NULL) %>%
dplyr::mutate_at("X1",
~ .x %>%
stringr::str_replace("\\.[0-9]$", "")) %>%
dplyr::semi_join(
sig_gene_comb,
by = c("X1" = "id")
) %>%
`colnames<-`(c("id", "class", "annotation", "function", "detail"))
head(gene_desc[c("id", "annotation")])
#> # A tibble: 6 x 2
#> id annotation
#> <chr> <chr>
#> 1 AT1G49860 glutathione S-transferase (class phi) 14
#> 2 AT1G07150 mitogen-activated protein kinase kinase kinase 13
#> 3 AT1G22500 RING/U-box superfamily protein
#> 4 AT1G49450 Transducin/WD40 repeat-like superfamily protein
#> 5 AT1G19050 response regulator 7
#> 6 AT1G14340 RNA-binding (RRM/RBD/RNP motifs) family proteinGo-term enrichment analysis was first performed with the significant genes from filtered count matrix.
go_params <- new(
"GOHyperGParams",
geneIds = sig_gene_comb[["id"]],
universeGeneIds = rownames(sam_count_filter),
annotation = "org.At.tair",
ontology = "BP",
pvalueCutoff = 0.001,
conditional = TRUE,
testDirection = "over"
)
print(go_params)
#> A GOHyperGParams instance
#> category: GO
#> annotation: org.At.tair
over_rep <- Category::hyperGTest(go_params)
print(summary(over_rep)[, c(1L, 2L, 5L, 6L, 7L)])
#> GOBPID Pvalue Count Size
#> 1 GO:0010167 1.596988e-06 6 18
#> 2 GO:0042221 1.810717e-05 64 1774
#> 3 GO:0009051 4.358056e-05 4 10
#> 4 GO:0006739 4.983977e-05 6 31
#> 5 GO:0006820 6.187222e-05 13 163
#> 6 GO:0042128 6.730251e-05 4 11
#> 7 GO:0055114 1.465420e-04 20 363
#> 8 GO:0019676 2.016793e-04 3 6
#> 9 GO:2001057 2.597026e-04 4 15
#> 10 GO:0009735 4.465533e-04 7 63
#> 11 GO:0046942 4.921617e-04 7 64
#> 12 GO:0006082 5.310063e-04 29 688
#> 13 GO:0016052 5.379109e-04 9 106
#> 14 GO:0006091 6.283807e-04 13 206
#> 15 GO:0001666 8.751107e-04 12 187
#> Term
#> 1 response to nitrate
#> 2 response to chemical
#> 3 pentose-phosphate shunt, oxidative branch
#> 4 NADP metabolic process
#> 5 anion transport
#> 6 nitrate assimilation
#> 7 oxidation-reduction process
#> 8 ammonia assimilation cycle
#> 9 reactive nitrogen species metabolic process
#> 10 response to cytokinin
#> 11 carboxylic acid transport
#> 12 organic acid metabolic process
#> 13 carbohydrate catabolic process
#> 14 generation of precursor metabolites and energy
#> 15 response to hypoxiaThe rlog-normalized count matrix (significant genes only) was
visualized with heatmap.
sig_heatmap_data <- sam_count_filter %>%
{
`rownames<-`(
DESeq2::rlog(., blind = FALSE),
rownames(.)
)
} %>%
`[`(sig_gene_comb[["id"]], 1:ncol(.))
sig_heatmap <- sig_heatmap_data %>%
{
`colnames<-`(
apply(., MARGIN = 1L, scale) %>%
t(),
colnames(.)
)
} %>%
ComplexHeatmap::Heatmap(
col = circlize::colorRamp2(c(-2, 0, 2),
c("#00FF00", "#000000", "#FF0000")),
name = "Value",
cluster_rows = TRUE,
cluster_columns = TRUE,
clustering_method_rows = "average",
clustering_method_columns = "average",
clustering_distance_rows = function(x) {
as.dist(1 - cor(t(x), method = "pearson"))
},
clustering_distance_columns = function(x) {
as.dist(1 - cor(t(x), method = "pearson"))
},
row_names_gp = grid::gpar(fontsize = 4),
column_names_gp = grid::gpar(fontsize = 8),
column_names_rot = 0,
column_names_centered = TRUE,
heatmap_width = grid::unit(1, "native"),
heatmap_height = grid::unit(2, "native")
)
ComplexHeatmap::draw(sig_heatmap)
In this workflow, we first prepared RNA-seq data from aligned “bam” files and turned the dataset into a count matrix with filtered gene reads. Then differentiating genes were discovered by univariate analysis with adjusted p-values and fold changes. Finally, significant genes went through GO enrichment analysis and the results were visualized with heatmaps.
This file was compiled with the following packages and versions:
utils::sessionInfo()
#> R version 4.0.5 (2021-03-31)
#> Platform: x86_64-w64-mingw32/x64 (64-bit)
#> Running under: Windows 10 x64 (build 19042)
#>
#> Matrix products: default
#>
#> locale:
#> [1] LC_COLLATE=Chinese (Simplified)_China.936
#> [2] LC_CTYPE=Chinese (Simplified)_China.936
#> [3] LC_MONETARY=Chinese (Simplified)_China.936
#> [4] LC_NUMERIC=C
#> [5] LC_TIME=Chinese (Simplified)_China.936
#>
#> attached base packages:
#> [1] grid stats4 parallel stats graphics grDevices utils
#> [8] datasets methods base
#>
#> other attached packages:
#> [1] ComplexHeatmap_2.4.3 org.At.tair.db_3.11.4
#> [3] GO.db_3.11.4 GOstats_2.54.0
#> [5] graph_1.66.0 Category_2.54.0
#> [7] Matrix_1.3-2 DESeq2_1.28.1
#> [9] DESeq_1.39.0 lattice_0.20-41
#> [11] locfit_1.5-9.4 GenomicAlignments_1.24.0
#> [13] SummarizedExperiment_1.18.2 DelayedArray_0.14.1
#> [15] matrixStats_0.56.0 GenomicFeatures_1.40.1
#> [17] AnnotationDbi_1.50.3 Biobase_2.48.0
#> [19] Rsamtools_2.4.0 Biostrings_2.56.0
#> [21] XVector_0.28.0 GenomicRanges_1.40.0
#> [23] GenomeInfoDb_1.24.2 IRanges_2.22.2
#> [25] S4Vectors_0.26.1 BiocGenerics_0.34.0
#> [27] rlang_0.4.11 magrittr_2.0.1
#> [29] forcats_0.5.1 stringr_1.4.0
#> [31] dplyr_1.0.7 purrr_0.3.4
#> [33] readr_2.0.1 tidyr_1.1.3
#> [35] tibble_3.1.3 ggplot2_3.3.5
#> [37] tidyverse_1.3.1
#>
#> loaded via a namespace (and not attached):
#> [1] colorspace_1.4-1 rjson_0.2.20 ellipsis_0.3.2
#> [4] circlize_0.4.13 GlobalOptions_0.1.2 fs_1.5.0
#> [7] clue_0.3-60 rstudioapi_0.13 bit64_4.0.5
#> [10] fansi_0.4.2 lubridate_1.7.10 xml2_1.3.2
#> [13] splines_4.0.5 cachem_1.0.4 geneplotter_1.66.0
#> [16] knitr_1.29 jsonlite_1.7.2 broom_0.7.9
#> [19] annotate_1.66.0 cluster_2.1.1 dbplyr_2.1.1
#> [22] png_0.1-7 compiler_4.0.5 httr_1.4.2
#> [25] backports_1.1.8 assertthat_0.2.1 fastmap_1.0.1
#> [28] cli_3.0.1 htmltools_0.5.0 prettyunits_1.1.1
#> [31] tools_4.0.5 gtable_0.3.0 glue_1.4.2
#> [34] GenomeInfoDbData_1.2.3 rappdirs_0.3.3 Rcpp_1.0.7
#> [37] cellranger_1.1.0 vctrs_0.3.8 rtracklayer_1.48.0
#> [40] xfun_0.15 rvest_1.0.1 lifecycle_1.0.0
#> [43] XML_3.99-0.4 zlibbioc_1.34.0 scales_1.1.1
#> [46] vroom_1.5.4 hms_1.1.0 RBGL_1.64.0
#> [49] RColorBrewer_1.1-2 yaml_2.2.1 curl_4.3
#> [52] memoise_2.0.0 biomaRt_2.44.4 stringi_1.4.6
#> [55] RSQLite_2.2.8 genefilter_1.70.0 BiocParallel_1.22.0
#> [58] shape_1.4.5 pkgconfig_2.0.3 bitops_1.0-6
#> [61] evaluate_0.14 bit_4.0.4 tidyselect_1.1.0
#> [64] AnnotationForge_1.30.1 GSEABase_1.50.1 R6_2.4.1
#> [67] snow_0.4-3 generics_0.1.0 DBI_1.1.0
#> [70] pillar_1.6.2 haven_2.4.3 withr_2.4.1
#> [73] survival_3.2-10 RCurl_1.98-1.2 modelr_0.1.8
#> [76] crayon_1.4.1 utf8_1.1.4 BiocFileCache_1.12.1
#> [79] tzdb_0.1.2 rmarkdown_2.3 GetoptLong_1.0.5
#> [82] progress_1.2.2 readxl_1.3.1 Rgraphviz_2.32.0
#> [85] blob_1.2.1 reprex_2.0.1 digest_0.6.25
#> [88] xtable_1.8-4 openssl_1.4.2 munsell_0.5.0
#> [91] askpass_1.1