Tutorial: RNA-seq differential expression & pathway analysis with Sailfish, DESeq2, GAGE, and Pathview

December 4, 2015
By

(This article was first published on Getting Genetics Done, and kindly contributed to R-bloggers)

Background

This tutorial shows an example of RNA-seq data analysis with DESeq2, followed by KEGG pathway analysis using GAGE. Using data from GSE37704, with processed data available on Figshare DOI: 10.6084/m9.figshare.1601975. This dataset has six samples from GSE37704, where expression was quantified by either: (A) mapping to to GRCh38 using STAR then counting reads mapped to genes with featureCounts under the union-intersection model, or (B) alignment-free quantification using Sailfish, summarized at the gene level using the GRCh38 GTF file. Both datasets are restricted to protein-coding genes only. Here I’ll use the Sailfish gene-level estimated counts.

Differential expression analysis

First, import the countdata and metadata directly from the web. Set up the DESeqDataSet, run the DESeq2 pipeline.
# Note importing BioC pkgs after dplyr requires explicitly using dplyr::select()
library(dplyr)
library(DESeq2)

# Which data do you want to use? Let's use the sailfish counts.
# browseURL("http://dx.doi.org/10.6084/m9.figshare.1601975")
# countDataURL = "http://files.figshare.com/2439061/GSE37704_featurecounts.csv"
countDataURL = "http://files.figshare.com/2600373/GSE37704_sailfish_genecounts.csv"

# Import countdata
countData = read.csv(countDataURL, row.names=1) %>%
dplyr::select(-length) %>%
as.matrix()

# Filter data where you only have 0 or 1 read count across all samples.
countData = countData[rowSums(countData)>1, ]
head(countData)
##                 SRR493366 SRR493367 SRR493368 SRR493369 SRR493370
## ENSG00000198888 17528 23007 30241 24418 29152
## ENSG00000198763 21264 26720 35550 28878 32416
## ENSG00000198804 130975 151207 195514 178130 196727
## ENSG00000198712 49769 61906 78608 66478 69758
## ENSG00000228253 9304 11160 12830 12608 13041
## ENSG00000198899 45401 51260 66851 63433 66123
## SRR493371
## ENSG00000198888 34416
## ENSG00000198763 38422
## ENSG00000198804 244670
## ENSG00000198712 86808
## ENSG00000228253 16063
## ENSG00000198899 79215
# Import metadata
colData = read.csv("http://files.figshare.com/2439060/GSE37704_metadata.csv", row.names=1)
colData
##               condition
## SRR493366 control_sirna
## SRR493367 control_sirna
## SRR493368 control_sirna
## SRR493369 hoxa1_kd
## SRR493370 hoxa1_kd
## SRR493371 hoxa1_kd
# Set up the DESeqDataSet Object and run the DESeq pipeline
dds = DESeqDataSetFromMatrix(countData=countData,
colData=colData,
design=~condition)
dds = DESeq(dds)
dds
## class: DESeqDataSet 
## dim: 16755 6
## metadata(0):
## assays(3): counts mu cooks
## rownames(16755): ENSG00000198888 ENSG00000198763 ...
## ENSG00000267795 ENSG00000165795
## rowRanges metadata column names(27): baseMean baseVar ... deviance
## maxCooks
## colnames(6): SRR493366 SRR493367 ... SRR493370 SRR493371
## colData names(2): condition sizeFactor
Next, get results for the HoxA1 knockdown versus control siRNA, and reorder them by p-value. Call summary on the results object to get a sense of how many genes are up or down-regulated at FDR 0.1.
res = results(dds, contrast=c("condition", "hoxa1_kd", "control_sirna"))
res = res[order(res$pvalue),]
summary(res)
## 
## out of 16755 with nonzero total read count
## adjusted p-value < 0.1
## LFC > 0 (up) : 4193, 25%
## LFC < 0 (down) : 4286, 26%
## outliers [1] : 22, 0.13%
## low counts [2] : 1299, 7.8%
## (mean count < 1)
## [1] see 'cooksCutoff' argument of ?results
## [2] see 'independentFiltering' argument of ?results
Since we mapped and counted against the Ensembl annotation, our results only have information about Ensembl gene IDs. But, our pathway analysis downstream will use KEGG pathways, and genes in KEGG pathways are annotated with Entrez gene IDs. I wrote an R package for doing this offline the dplyr way (https://github.com/stephenturner/annotables), but the canonical Bioconductor way to do it is with the AnnotationDbi and organism annotation packages. Here we’re using the organism package (“org”) for Homo sapiens (“Hs”), organized as an AnnotationDbi database package (“db”) using Entrez Gene IDs (“eg”) as primary keys. To see what all the keys are, use the columns function.
library("AnnotationDbi")
library("org.Hs.eg.db")
columns(org.Hs.eg.db)
##  [1] "ACCNUM"       "ALIAS"        "ENSEMBL"      "ENSEMBLPROT" 
## [5] "ENSEMBLTRANS" "ENTREZID" "ENZYME" "EVIDENCE"
## [9] "EVIDENCEALL" "GENENAME" "GO" "GOALL"
## [13] "IPI" "MAP" "OMIM" "ONTOLOGY"
## [17] "ONTOLOGYALL" "PATH" "PFAM" "PMID"
## [21] "PROSITE" "REFSEQ" "SYMBOL" "UCSCKG"
## [25] "UNIGENE" "UNIPROT"
Let’s use the mapIds function to add more columns to the results. The row.names of our results table has the Ensembl gene ID (our key), so we need to specify keytype=ENSEMBL. The column argument tells the mapIds function which information we want, and the multiVals argument tells the function what to do if there are multiple possible values for a single input value. Here we ask to just give us back the first one that occurs in the database. Let’s get the Entrez IDs, gene symbols, and full gene names.
res$symbol = mapIds(org.Hs.eg.db,
keys=row.names(res),
column="SYMBOL",
keytype="ENSEMBL",
multiVals="first")
res$entrez = mapIds(org.Hs.eg.db,
keys=row.names(res),
column="ENTREZID",
keytype="ENSEMBL",
multiVals="first")
res$name = mapIds(org.Hs.eg.db,
keys=row.names(res),
column="GENENAME",
keytype="ENSEMBL",
multiVals="first")

head(res, 10)
## log2 fold change (MAP): condition hoxa1_kd vs control_sirna 
## Wald test p-value: condition hoxa1_kd vs control_sirna
## DataFrame with 10 rows and 9 columns
## baseMean log2FoldChange lfcSE stat pvalue
##
## ENSG00000148773 1885.344 -3.172502 0.07868572 -40.31865 0
## ENSG00000138623 2939.936 -2.418238 0.05889229 -41.06205 0
## ENSG00000104368 13601.963 2.016802 0.05249643 38.41789 0
## ENSG00000124766 2692.200 2.379545 0.06193654 38.41908 0
## ENSG00000122861 35889.413 2.224779 0.05258658 42.30697 0
## ENSG00000116016 4558.157 -1.885339 0.04258766 -44.26961 0
## ENSG00000164251 2404.103 3.325196 0.07021236 47.35912 0
## ENSG00000125257 6187.386 1.943762 0.04259189 45.63692 0
## ENSG00000104321 9334.555 3.186856 0.06227530 51.17367 0
## ENSG00000183508 2110.345 3.190612 0.07488305 42.60794 0
## padj symbol entrez
##
## ENSG00000148773 0 MKI67 4288
## ENSG00000138623 0 SEMA7A 8482
## ENSG00000104368 0 PLAT 5327
## ENSG00000124766 0 SOX4 6659
## ENSG00000122861 0 PLAU 5328
## ENSG00000116016 0 EPAS1 2034
## ENSG00000164251 0 F2RL1 2150
## ENSG00000125257 0 ABCC4 10257
## ENSG00000104321 0 TRPA1 8989
## ENSG00000183508 0 FAM46C 54855
## name
##
## ENSG00000148773 marker of proliferation Ki-67
## ENSG00000138623 semaphorin 7A, GPI membrane anchor (John Milton Hagen blood group)
## ENSG00000104368 plasminogen activator, tissue
## ENSG00000124766 SRY (sex determining region Y)-box 4
## ENSG00000122861 plasminogen activator, urokinase
## ENSG00000116016 endothelial PAS domain protein 1
## ENSG00000164251 coagulation factor II (thrombin) receptor-like 1
## ENSG00000125257 ATP-binding cassette, sub-family C (CFTR/MRP), member 4
## ENSG00000104321 transient receptor potential cation channel, subfamily A, member 1
## ENSG00000183508 family with sequence similarity 46, member C

Pathway analysis

We’re going to use the gage package (Generally Applicable Gene-set Enrichment for Pathway Analysis) for pathway analysis. See also the gage package workflow vignette for RNA-seq pathway analysis. Once we have a list of enriched pathways, we’re going to use the pathview package to draw pathway diagrams, shading the molecules in the pathway by their degree of up/down-regulation.

KEGG pathways

The gageData package has pre-compiled databases mapping genes to KEGG pathways and GO terms for common organisms. kegg.sets.hs is a named list of 229 elements. Each element is a character vector of member gene Entrez IDs for a single KEGG pathway. (See also go.sets.hs). sigmet.idx.hs is an index of numbers of sinaling and metabolic pathways in kegg.set.gs. In other words, KEGG pathway include other types of pathway definitions, like “Global Map” and “Human Diseases”, which may be undesirable in pathway analysis. Therefore, kegg.sets.hs[sigmet.idx.hs] gives you the “cleaner” gene sets of sinaling and metabolic pathways only.
library(pathview)
library(gage)
library(gageData)
data(kegg.sets.hs)
data(sigmet.idx.hs)
kegg.sets.hs = kegg.sets.hs[sigmet.idx.hs]
head(kegg.sets.hs, 3)
## $`hsa00232 Caffeine metabolism`
## [1] "10" "1544" "1548" "1549" "1553" "7498" "9"
##
## $`hsa00983 Drug metabolism - other enzymes`
## [1] "10" "1066" "10720" "10941" "151531" "1548" "1549"
## [8] "1551" "1553" "1576" "1577" "1806" "1807" "1890"
## [15] "221223" "2990" "3251" "3614" "3615" "3704" "51733"
## [22] "54490" "54575" "54576" "54577" "54578" "54579" "54600"
## [29] "54657" "54658" "54659" "54963" "574537" "64816" "7083"
## [36] "7084" "7172" "7363" "7364" "7365" "7366" "7367"
## [43] "7371" "7372" "7378" "7498" "79799" "83549" "8824"
## [50] "8833" "9" "978"
##
## $`hsa00230 Purine metabolism`
## [1] "100" "10201" "10606" "10621" "10622" "10623" "107"
## [8] "10714" "108" "10846" "109" "111" "11128" "11164"
## [15] "112" "113" "114" "115" "122481" "122622" "124583"
## [22] "132" "158" "159" "1633" "171568" "1716" "196883"
## [29] "203" "204" "205" "221823" "2272" "22978" "23649"
## [36] "246721" "25885" "2618" "26289" "270" "271" "27115"
## [43] "272" "2766" "2977" "2982" "2983" "2984" "2986"
## [50] "2987" "29922" "3000" "30833" "30834" "318" "3251"
## [57] "353" "3614" "3615" "3704" "377841" "471" "4830"
## [64] "4831" "4832" "4833" "4860" "4881" "4882" "4907"
## [71] "50484" "50940" "51082" "51251" "51292" "5136" "5137"
## [78] "5138" "5139" "5140" "5141" "5142" "5143" "5144"
## [85] "5145" "5146" "5147" "5148" "5149" "5150" "5151"
## [92] "5152" "5153" "5158" "5167" "5169" "51728" "5198"
## [99] "5236" "5313" "5315" "53343" "54107" "5422" "5424"
## [106] "5425" "5426" "5427" "5430" "5431" "5432" "5433"
## [113] "5434" "5435" "5436" "5437" "5438" "5439" "5440"
## [120] "5441" "5471" "548644" "55276" "5557" "5558" "55703"
## [127] "55811" "55821" "5631" "5634" "56655" "56953" "56985"
## [134] "57804" "58497" "6240" "6241" "64425" "646625" "654364"
## [141] "661" "7498" "8382" "84172" "84265" "84284" "84618"
## [148] "8622" "8654" "87178" "8833" "9060" "9061" "93034"
## [155] "953" "9533" "954" "955" "956" "957" "9583"
## [162] "9615"
The gage() function requires a named vector of fold changes, where the names of the values are the Entrez gene IDs.
foldchanges = res$log2FoldChange
names(foldchanges) = res$entrez
head(foldchanges)
##      4288      8482      5327      6659      5328      2034 
## -3.172502 -2.418238 2.016802 2.379545 2.224779 -1.885339
Now, let’s run the pathway analysis. See help on the gage function with ?gage. Specifically, you might want to try changing the value of same.dir. This value determins whether to test for changes in a gene set toward a single direction (all genes up or down regulated) or changes towards both directions simultaneously (any genes in the pathway dysregulated).
For experimentally derived gene sets, GO term groups, etc, coregulation is commonly the case, hence same.dir = TRUE (default); In KEGG, BioCarta pathways, genes frequently are not coregulated, hence it could be informative to let same.dir = FALSE. Although same.dir = TRUE could also be interesting for pathways.
Here, we’re using same.dir = TRUE, which will give us separate lists for pathways that are upregulated versus pathways that are downregulated. Let’s look at the first few results from each.
# Get the results
keggres = gage(foldchanges, gsets=kegg.sets.hs, same.dir=TRUE)

# Look at both up (greater), down (less), and statatistics.
lapply(keggres, head)
## $greater
## p.geomean stat.mean p.val
## hsa04142 Lysosome 0.0002630657 3.517890 0.0002630657
## hsa04640 Hematopoietic cell lineage 0.0017919390 2.976432 0.0017919390
## hsa04630 Jak-STAT signaling pathway 0.0048980977 2.604390 0.0048980977
## hsa00140 Steroid hormone biosynthesis 0.0051115493 2.636206 0.0051115493
## hsa04062 Chemokine signaling pathway 0.0125582961 2.250765 0.0125582961
## hsa00511 Other glycan degradation 0.0223819919 2.104311 0.0223819919
## q.val set.size exp1
## hsa04142 Lysosome 0.04261664 116 0.0002630657
## hsa04640 Hematopoietic cell lineage 0.14514706 61 0.0017919390
## hsa04630 Jak-STAT signaling pathway 0.20701775 119 0.0048980977
## hsa00140 Steroid hormone biosynthesis 0.20701775 39 0.0051115493
## hsa04062 Chemokine signaling pathway 0.40688879 156 0.0125582961
## hsa00511 Other glycan degradation 0.49956506 15 0.0223819919
##
## $less
## p.geomean stat.mean p.val
## hsa04110 Cell cycle 2.165725e-06 -4.722301 2.165725e-06
## hsa03030 DNA replication 3.807440e-06 -4.835336 3.807440e-06
## hsa04114 Oocyte meiosis 1.109869e-04 -3.767561 1.109869e-04
## hsa03013 RNA transport 1.181787e-03 -3.071947 1.181787e-03
## hsa03440 Homologous recombination 1.197124e-03 -3.190747 1.197124e-03
## hsa00240 Pyrimidine metabolism 1.570318e-03 -2.992059 1.570318e-03
## q.val set.size exp1
## hsa04110 Cell cycle 0.0003084027 121 2.165725e-06
## hsa03030 DNA replication 0.0003084027 36 3.807440e-06
## hsa04114 Oocyte meiosis 0.0059932916 101 1.109869e-04
## hsa03013 RNA transport 0.0387868193 145 1.181787e-03
## hsa03440 Homologous recombination 0.0387868193 28 1.197124e-03
## hsa00240 Pyrimidine metabolism 0.0423985796 96 1.570318e-03
##
## $stats
## stat.mean exp1
## hsa04142 Lysosome 3.517890 3.517890
## hsa04640 Hematopoietic cell lineage 2.976432 2.976432
## hsa04630 Jak-STAT signaling pathway 2.604390 2.604390
## hsa00140 Steroid hormone biosynthesis 2.636206 2.636206
## hsa04062 Chemokine signaling pathway 2.250765 2.250765
## hsa00511 Other glycan degradation 2.104311 2.104311
Now, let’s process the results to pull out the top 5 upregulated pathways, then further process that just to get the IDs. We’ll use these KEGG pathway IDs downstream for plotting.
# Get the pathways
keggrespathways = data.frame(id=rownames(keggres$greater), keggres$greater) %>%
tbl_df() %>%
filter(row_number()<=5) %>%
.$id %>%
as.character()
keggrespathways
## [1] "hsa04142 Lysosome"                    
## [2] "hsa04640 Hematopoietic cell lineage"
## [3] "hsa04630 Jak-STAT signaling pathway"
## [4] "hsa00140 Steroid hormone biosynthesis"
## [5] "hsa04062 Chemokine signaling pathway"
# Get the IDs.
keggresids = substr(keggrespathways, start=1, stop=8)
keggresids
## [1] "hsa04142" "hsa04640" "hsa04630" "hsa00140" "hsa04062"
Finally, the pathview() function in the pathview package makes the plots. Let’s write a function so we can loop through and draw plots for the top 5 pathways we created above.
# Define plotting function for applying later
plot_pathway = function(pid) pathview(gene.data=foldchanges, pathway.id=pid, species="hsa", new.signature=FALSE)

# plot multiple pathways (plots saved to disk and returns a throwaway list object)
tmp = sapply(keggresids, function(pid) pathview(gene.data=foldchanges, pathway.id=pid, species="hsa"))
Here are the plots:

Gene Ontology (GO)

We can also do a similar procedure with gene ontology. Similar to above, go.sets.hs has all GO terms. go.subs.hs is a named list containing indexes for the BP, CC, and MF ontologies. Let’s only do Biological Process.
data(go.sets.hs)
data(go.subs.hs)
gobpsets = go.sets.hs[go.subs.hs$BP]

gobpres = gage(foldchanges, gsets=gobpsets, same.dir=TRUE)

lapply(gobpres, head)
## $greater
## p.geomean
## GO:0007156 homophilic cell adhesion 3.914568e-05
## GO:0008285 negative regulation of cell proliferation 2.907332e-04
## GO:0016339 calcium-dependent cell-cell adhesion 4.218753e-04
## GO:0016337 cell-cell adhesion 6.170551e-04
## GO:0048729 tissue morphogenesis 6.581460e-04
## GO:1901617 organic hydroxy compound biosynthetic process 8.876161e-04
## stat.mean
## GO:0007156 homophilic cell adhesion 4.017207
## GO:0008285 negative regulation of cell proliferation 3.453345
## GO:0016339 calcium-dependent cell-cell adhesion 3.543891
## GO:0016337 cell-cell adhesion 3.244296
## GO:0048729 tissue morphogenesis 3.223979
## GO:1901617 organic hydroxy compound biosynthetic process 3.157421
## p.val
## GO:0007156 homophilic cell adhesion 3.914568e-05
## GO:0008285 negative regulation of cell proliferation 2.907332e-04
## GO:0016339 calcium-dependent cell-cell adhesion 4.218753e-04
## GO:0016337 cell-cell adhesion 6.170551e-04
## GO:0048729 tissue morphogenesis 6.581460e-04
## GO:1901617 organic hydroxy compound biosynthetic process 8.876161e-04
## q.val
## GO:0007156 homophilic cell adhesion 0.1613977
## GO:0008285 negative regulation of cell proliferation 0.4720349
## GO:0016339 calcium-dependent cell-cell adhesion 0.4720349
## GO:0016337 cell-cell adhesion 0.4720349
## GO:0048729 tissue morphogenesis 0.4720349
## GO:1901617 organic hydroxy compound biosynthetic process 0.4720349
## set.size
## GO:0007156 homophilic cell adhesion 124
## GO:0008285 negative regulation of cell proliferation 458
## GO:0016339 calcium-dependent cell-cell adhesion 27
## GO:0016337 cell-cell adhesion 355
## GO:0048729 tissue morphogenesis 429
## GO:1901617 organic hydroxy compound biosynthetic process 141
## exp1
## GO:0007156 homophilic cell adhesion 3.914568e-05
## GO:0008285 negative regulation of cell proliferation 2.907332e-04
## GO:0016339 calcium-dependent cell-cell adhesion 4.218753e-04
## GO:0016337 cell-cell adhesion 6.170551e-04
## GO:0048729 tissue morphogenesis 6.581460e-04
## GO:1901617 organic hydroxy compound biosynthetic process 8.876161e-04
##
## $less
## p.geomean stat.mean
## GO:0048285 organelle fission 4.411540e-18 -8.850004
## GO:0000280 nuclear division 7.459684e-18 -8.805564
## GO:0007067 mitosis 7.459684e-18 -8.805564
## GO:0000087 M phase of mitotic cell cycle 2.286444e-17 -8.655644
## GO:0007059 chromosome segregation 1.872901e-13 -7.686883
## GO:0051301 cell division 5.841375e-12 -6.887763
## p.val q.val
## GO:0048285 organelle fission 4.411540e-18 1.025209e-14
## GO:0000280 nuclear division 7.459684e-18 1.025209e-14
## GO:0007067 mitosis 7.459684e-18 1.025209e-14
## GO:0000087 M phase of mitotic cell cycle 2.286444e-17 2.356752e-14
## GO:0007059 chromosome segregation 1.872901e-13 1.544394e-10
## GO:0051301 cell division 5.841375e-12 4.013998e-09
## set.size exp1
## GO:0048285 organelle fission 376 4.411540e-18
## GO:0000280 nuclear division 352 7.459684e-18
## GO:0007067 mitosis 352 7.459684e-18
## GO:0000087 M phase of mitotic cell cycle 362 2.286444e-17
## GO:0007059 chromosome segregation 141 1.872901e-13
## GO:0051301 cell division 462 5.841375e-12
##
## $stats
## stat.mean
## GO:0007156 homophilic cell adhesion 4.017207
## GO:0008285 negative regulation of cell proliferation 3.453345
## GO:0016339 calcium-dependent cell-cell adhesion 3.543891
## GO:0016337 cell-cell adhesion 3.244296
## GO:0048729 tissue morphogenesis 3.223979
## GO:1901617 organic hydroxy compound biosynthetic process 3.157421
## exp1
## GO:0007156 homophilic cell adhesion 4.017207
## GO:0008285 negative regulation of cell proliferation 3.453345
## GO:0016339 calcium-dependent cell-cell adhesion 3.543891
## GO:0016337 cell-cell adhesion 3.244296
## GO:0048729 tissue morphogenesis 3.223979
## GO:1901617 organic hydroxy compound biosynthetic process 3.157421

To leave a comment for the author, please follow the link and comment on their blog: Getting Genetics Done.

R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: Data science, Big Data, R jobs, visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.

Search R-bloggers


Sponsors

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)