![QC
and
pre-‐processing
• First
step
in
QC:
– Look
at
quality
scores
to
see
if
sequencing
was
successful
• Sequence
data
usually
stored
in
FASTQ
format:
@BI:080831_SL-XAN_0004_30BV1AAXX:8:1:731:1429#0/1
GTTTCAACGGGTGTTGGAATCCACACCAAACAATGGCTACCTCTATCACCC
+
hbhhP_Z[`VFhHNU]KTWPHHIKMIIJKDJGGJGEDECDCGCABEAFEB
Header
(typically
w/
flowcell
#)
Sequence
Quality
scores
flow
cell
lane
Cle
number
x-‐coordinate
y-‐coordinate
provided
by
user
1st
end
of
paired
read
40,34,40,40,16,31,26,28,27,32,22,6,40,8,14,21,29,11,20,23,16,…
ASCII
table
Numerical
quality
scores
Typical
range
of
quality
scores:
0
~
40](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-5-320.jpg)
![Tutorial:
Installing
BioConductor
packages
source("http://bioconductor.org/biocLite.R")
biocLite("DESeq2")
hep://www.bioconductor.org/
M.
I.
Love,
W.
Huber,
S.
Anders:
Moderated
esCmaCon
of
fold
change
and
dispersion
for
RNA-‐Seq
data
with
DESeq2.
bioRxiv
(2014).
doi:
10.1101/002832
[1]](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-26-320.jpg)
![Tutorial:
DESeq2
analysis
# load library
library(DESeq2)
# create experiment labels (two conditions)
colData <- DataFrame(condition=factor(c("ctrl","ctrl", "ctrl", "treat", "treat", "treat")))
# create DESeq input matrix
dds <- DESeqDataSetFromMatrix(countData, colData, formula(~ condition))
# run DEseq
dds <- DESeq(dds)
# visualize differentially expressed genes
plotMA(dds)
# get differentially expressed genes
res <- results(dds)
# order by BH adjusted p-value
resOrdered <- res[order(res$padj),]
# top of ordered matrix
head(resOrdered)](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-29-320.jpg)
![Tutorial:
DESeq2
analysis
# get differentially expressed genes
res <- results(dds)
# order by BH adjusted p-value
resOrdered <- res[order(res$padj),]
# top of ordered matrix
head(resOrdered)
DataFrame with 6 rows and 6 columns
baseMean log2FoldChange lfcSE stat pvalue padj
<numeric> <numeric> <numeric> <numeric> <numeric> <numeric>
Pck1 19300.0081 -2.3329116 0.16519373 -14.12228 2.768978e-45 3.986497e-41
Fras1 1202.1842 -0.8469410 0.06499738 -13.03039 8.219001e-39 5.916448e-35
S100a14 590.6305 2.1903041 0.17608923 12.43860 1.612985e-35 7.740716e-32
Ugt1a2 2759.7012 -1.7037495 0.15339576 -11.10689 1.161372e-28 4.180067e-25
Crip1 681.0106 0.7717364 0.07264577 10.62328 2.322502e-26 5.572844e-23
Smpdl3a 11152.4458 0.3398371 0.03195000 10.63653 2.014913e-26 5.572844e-23
# how many differentially expressed genes ? FDR=10%, |fold-change|>2 (up and down)](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-30-320.jpg)
![Tutorial:
DESeq2
analysis
# how many differentially expressed genes ? FDR=10%, |fold-change|>2 (up and down)
# get differentially expressed gene matrix
sig <- resOrdered[!is.na(resOrdered$padj) &
resOrdered$padj<0.10 &
abs(resOrdered$log2FoldChange)>=1,]](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-31-320.jpg)
![Tutorial:
DESeq2
analysis
# how many differentially expressed genes ? FDR=10%, |fold-change|>2 (up and down)
# get differentially expressed gene matrix
sig <- resOrdered[!is.na(resOrdered$padj) &
resOrdered$padj<0.10 &
abs(resOrdered$log2FoldChange)>=1,]
head(sig)
DataFrame with 6 rows and 6 columns
baseMean log2FoldChange lfcSE stat pvalue padj
<numeric> <numeric> <numeric> <numeric> <numeric> <numeric>
Pck1 19300 -2.33 0.165 -14.12 2.77e-45 3.99e-41
S100a14 591 2.19 0.176 12.44 1.61e-35 7.74e-32
Ugt1a2 2760 -1.70 0.153 -11.11 1.16e-28 4.18e-25
Pklr 787 -1.00 0.097 -10.34 4.62e-25 9.49e-22
Mlph 1321 1.20 0.117 10.20 1.90e-24 3.42e-21
Ifit1 285 1.39 0.156 8.94 3.76e-19 3.38e-16
dim(sig)
# how to create a heat map](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-32-320.jpg)
![Tutorial:
Heat
Map
# how to create a heat map
# select genes
selected <- rownames(sig);selected
## load libraries for the heat map
library("RColorBrewer")
source("http://bioconductor.org/biocLite.R")
biocLite(”gplots”)
library("gplots")
# colors of the heat map
hmcol <- colorRampPalette(brewer.pal(9, "GnBu"))(100) ## hmcol <- heat.colors
heatmap.2( log2(counts(dds,normalized=TRUE)[rownames(dds) %in% selected,]),
col = hmcol, scale="row”,
Rowv = TRUE, Colv = FALSE,
dendrogram="row",
trace="none",
margin=c(4,6), cexRow=0.5, cexCol=1, keysize=1 )](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-33-320.jpg)
![Tutorial:
Heat
Map
# how to create a heat map
library("RColorBrewer")
library("gplots")
# colors of the heat map
hmcol <- colorRampPalette(brewer.pal(9, "GnBu"))(100) ## hmcol <- heat.colors
heatmap.2(log2(counts(dds,normalized=TRUE)[rownames(dds) %in% selected,]),
col = hmcol, Rowv = TRUE, Colv = FALSE, scale="row", dendrogram="row", trace="none",
margin=c(4,6), cexRow=0.5, cexCol=1, keysize=1 )](https://image.slidesharecdn.com/lecture20sboner-240405072050-d16a7392/85/RNA-sequencing-analysis-tutorial-with-NGS-34-320.jpg)
RNA sequencing analysis tutorial with NGS
- 1. RNA-‐seq data analysis tutorial Andrea Sboner 2015-‐05-‐21
- 2. NGS Experiment Data management: Mapping the reads CreaCng summaries Downstream analysis: the interes)ng stuff DifferenCal expression, chimeric transcripts, novel transcribed regions, etc.
- 3. What is RNA-‐seq? • Next-‐generaCon sequencing applied to the “transcriptome” ApplicaCons: Gene (exon, isoform) expression esCmaCon Differen)al gene (exon, isoform) expression analysis Discovery of novel transcribed regions Discovery/Detec)on of chimeric transcripts Allele specific expression …
- 4. NGS Experiment Data management: Mapping the reads CreaCng summaries Downstream analysis: the interes)ng stuff DifferenCal expression, chimeric transcripts, novel transcribed regions, etc.
- 5. QC and pre-‐processing • First step in QC: – Look at quality scores to see if sequencing was successful • Sequence data usually stored in FASTQ format: @BI:080831_SL-XAN_0004_30BV1AAXX:8:1:731:1429#0/1 GTTTCAACGGGTGTTGGAATCCACACCAAACAATGGCTACCTCTATCACCC + hbhhP_Z[`VFhHNU]KTWPHHIKMIIJKDJGGJGEDECDCGCABEAFEB Header (typically w/ flowcell #) Sequence Quality scores flow cell lane Cle number x-‐coordinate y-‐coordinate provided by user 1st end of paired read 40,34,40,40,16,31,26,28,27,32,22,6,40,8,14,21,29,11,20,23,16,… ASCII table Numerical quality scores Typical range of quality scores: 0 ~ 40
- 6. Freely available tools for QC • FastQC – hep://www.bioinformaCcs.bbsrc.ac.uk/projects/fastqc/ – Nice GUI and command line interface • FASTX-‐Toolkit – hep://hannonlab.cshl.edu/fastx_toolkit/index.html – Tools for QC as well as trimming reads, removing adapters, filtering by read quality, etc. • Galaxy – hep://main.g2.bx.psu.edu/ – Web interface – Many funcCons but analyses are done on remote server
- 7. FastQC • GUI mode fastqc • Command line mode fastqc fastq_files –o output_directory – will create fastq_file_fastqc.zip in output directory
- 8. FastQC read 1 rule of thumb: average quality > 20 for the first 36bp -‐ median -‐ mean
- 9. What to do when quality is poor? • Trim the reads • FASTX-‐toolkit – fastx_trimmer –f N –l N – fastq_quality_filter -‐q N –p N – Fastx_clipper -‐a ADAPTER
- 10. NGS Experiment Data management: Mapping the reads CreaCng summaries Downstream analysis: the interes)ng stuff DifferenCal expression, chimeric transcripts, novel transcribed regions, etc.
- 11. Mapping InsCtute for ComputaConal Biomedicine
- 13. Mapping to a reference • Genome • Transcriptome • Genome + Transcriptome • Transcriptome + Genome • Genome + splice juncCon library reference transcriptome
- 14. Alignment tools • BWA – hep://bio-‐bwa.sourceforge.net/bwa.shtml – Gapped alignments (good for indel detecCon) • BowCe – hep://bowCe-‐bio.sourceforge.net/index.shtml – Supports gapped alignments in latest version (bowCe 2) • TopHat – hep://tophat.cbcb.umd.edu/ – Good for discovering novel transcripts in RNA-‐seq data – Builds exon models and splice juncCons de novo. – Requires more CPU Cme and disk space • STAR – heps://code.google.com/p/rna-‐star/ – Detects splice juncCons de novo – Super fast: ~10min for 200M reads but – Requires 21Gb of memory • More than 70 short-‐read aligners: – hep://en.wikipedia.org/wiki/List_of_sequence_alignment_sooware
- 15. NGS Experiment Data management: Mapping the reads CreaCng summaries Downstream analysis: the interes)ng stuff DifferenCal expression, chimeric transcripts, novel transcribed regions, etc.
- 16. Analyzing RNA-‐Seq experiments • How many molecules of mRNA1 are in my sample? – EsCmaCng expression • Is the amount or mRNA1 in sample/group A different from sample/group B ? – DifferenCal analysis
- 17. Es)ma)ng expression: counCng how many RNA-‐seq reads map to genes • Using R – summarizeOverlaps in GenomicRanges – easyRNASeq • Using Python – htseq-‐count • How it works: – SAM/BAM files (TopHat2, STAR, …) – Gene annotaCon (GFF, GTF format)
- 18. GFF/GTF file format: hep://en.wikipedia.org/wiki/General_feature_format hep://useast.ensembl.org/info/website/upload/gff.html hep://www.sanger.ac.uk/resources/sooware/gff/ hep://www.sequenceontology.org/gff3.shtml
- 19. GFF/GTF file format: hep://en.wikipedia.org/wiki/General_feature_format hep://useast.ensembl.org/info/website/upload/gff.html hep://www.sanger.ac.uk/resources/sooware/gff/ hep://www.sequenceontology.org/gff3.shtml
- 20. Tutorial: RNA-‐seq count matrix • Download – hep://icb.med.cornell.edu/faculty/sboner/lab/ EpigenomicsWorkshop/count_matrix.txt • Load into R, inspect
- 21. Tutorial: RNA-‐seq count matrix # working directory getwd() # read in count matrix countData <- read.csv("count_matrix.txt", header=T, row.names=1, sep="t") dim(countData) head(countData)
- 22. Read counts GENE ctrl1 ctrl2 ctrl3 treat1 treat2 treat3 0610005C13Rik 1438 1104 1825 1348 1154 1005 0610007N19Rik 1012 1152 1139 878 885 835 0610007P14Rik 704 796 881 826 865 929 0610009B22Rik 757 802 780 885 853 987 0610009D07Rik 1107 1183 1220 1258 1221 1428 … … … … … … … 24009 rows, i.e. genes 6 columns, i.e. samples
- 23. Tutorial: Basic QC barplot(colSums(countData)*1e-6, names=colnames(countData), ylab="Library size (millions)")
- 24. Tutorial: Basic QC barplot(colSums(countData)*1e-6, names=colnames(countData), ylab="Library size (millions)")
- 25. Analyzing expression • How many molecules of mRNA1 are in my sample? – EsCmaCng expression • Is the amount or mRNA1 in sample/group A different from sample/group B ? – DifferenCal analysis
- 26. Tutorial: Installing BioConductor packages source("http://bioconductor.org/biocLite.R") biocLite("DESeq2") hep://www.bioconductor.org/ M. I. Love, W. Huber, S. Anders: Moderated esCmaCon of fold change and dispersion for RNA-‐Seq data with DESeq2. bioRxiv (2014). doi: 10.1101/002832 [1]
- 27. Tutorial: DESeq2 analysis # load library library(DESeq2) # create experiment labels (two conditions) colData <- DataFrame(condition=factor(c("ctrl","ctrl", "ctrl", "treat", "treat", "treat"))) # create DESeq input matrix dds <- DESeqDataSetFromMatrix(countData, colData, formula(~ condition)) # run DEseq dds <- DESeq(dds) # visualize differentially expressed genes plotMA(dds)
- 28. Tutorial: DESeq2 analysis # load library library(DESeq2) # create experiment labels (two conditions) colData <- DataFrame(condition=factor(c("ctrl","ctrl", "ctrl", "treat", "treat", "treat"))) # create DESeq input matrix dds <- DESeqDataSetFromMatrix(countData, colData, formula(~ condition)) # run DEseq dds <- DESeq(dds) # visualize differentially expressed genes plotMA(dds)
- 29. Tutorial: DESeq2 analysis # load library library(DESeq2) # create experiment labels (two conditions) colData <- DataFrame(condition=factor(c("ctrl","ctrl", "ctrl", "treat", "treat", "treat"))) # create DESeq input matrix dds <- DESeqDataSetFromMatrix(countData, colData, formula(~ condition)) # run DEseq dds <- DESeq(dds) # visualize differentially expressed genes plotMA(dds) # get differentially expressed genes res <- results(dds) # order by BH adjusted p-value resOrdered <- res[order(res$padj),] # top of ordered matrix head(resOrdered)
- 30. Tutorial: DESeq2 analysis # get differentially expressed genes res <- results(dds) # order by BH adjusted p-value resOrdered <- res[order(res$padj),] # top of ordered matrix head(resOrdered) DataFrame with 6 rows and 6 columns baseMean log2FoldChange lfcSE stat pvalue padj <numeric> <numeric> <numeric> <numeric> <numeric> <numeric> Pck1 19300.0081 -2.3329116 0.16519373 -14.12228 2.768978e-45 3.986497e-41 Fras1 1202.1842 -0.8469410 0.06499738 -13.03039 8.219001e-39 5.916448e-35 S100a14 590.6305 2.1903041 0.17608923 12.43860 1.612985e-35 7.740716e-32 Ugt1a2 2759.7012 -1.7037495 0.15339576 -11.10689 1.161372e-28 4.180067e-25 Crip1 681.0106 0.7717364 0.07264577 10.62328 2.322502e-26 5.572844e-23 Smpdl3a 11152.4458 0.3398371 0.03195000 10.63653 2.014913e-26 5.572844e-23 # how many differentially expressed genes ? FDR=10%, |fold-change|>2 (up and down)
- 31. Tutorial: DESeq2 analysis # how many differentially expressed genes ? FDR=10%, |fold-change|>2 (up and down) # get differentially expressed gene matrix sig <- resOrdered[!is.na(resOrdered$padj) & resOrdered$padj<0.10 & abs(resOrdered$log2FoldChange)>=1,]
- 32. Tutorial: DESeq2 analysis # how many differentially expressed genes ? FDR=10%, |fold-change|>2 (up and down) # get differentially expressed gene matrix sig <- resOrdered[!is.na(resOrdered$padj) & resOrdered$padj<0.10 & abs(resOrdered$log2FoldChange)>=1,] head(sig) DataFrame with 6 rows and 6 columns baseMean log2FoldChange lfcSE stat pvalue padj <numeric> <numeric> <numeric> <numeric> <numeric> <numeric> Pck1 19300 -2.33 0.165 -14.12 2.77e-45 3.99e-41 S100a14 591 2.19 0.176 12.44 1.61e-35 7.74e-32 Ugt1a2 2760 -1.70 0.153 -11.11 1.16e-28 4.18e-25 Pklr 787 -1.00 0.097 -10.34 4.62e-25 9.49e-22 Mlph 1321 1.20 0.117 10.20 1.90e-24 3.42e-21 Ifit1 285 1.39 0.156 8.94 3.76e-19 3.38e-16 dim(sig) # how to create a heat map
- 33. Tutorial: Heat Map # how to create a heat map # select genes selected <- rownames(sig);selected ## load libraries for the heat map library("RColorBrewer") source("http://bioconductor.org/biocLite.R") biocLite(”gplots”) library("gplots") # colors of the heat map hmcol <- colorRampPalette(brewer.pal(9, "GnBu"))(100) ## hmcol <- heat.colors heatmap.2( log2(counts(dds,normalized=TRUE)[rownames(dds) %in% selected,]), col = hmcol, scale="row”, Rowv = TRUE, Colv = FALSE, dendrogram="row", trace="none", margin=c(4,6), cexRow=0.5, cexCol=1, keysize=1 )
- 34. Tutorial: Heat Map # how to create a heat map library("RColorBrewer") library("gplots") # colors of the heat map hmcol <- colorRampPalette(brewer.pal(9, "GnBu"))(100) ## hmcol <- heat.colors heatmap.2(log2(counts(dds,normalized=TRUE)[rownames(dds) %in% selected,]), col = hmcol, Rowv = TRUE, Colv = FALSE, scale="row", dendrogram="row", trace="none", margin=c(4,6), cexRow=0.5, cexCol=1, keysize=1 )
- 35. SelecCng the most differenCally expressed genes and run GO analysis # universe universe <- rownames(resOrdered) # load mouse annotation and ID library biocLite(“org.Mm.eg.db”) library(org.Mm.eg.db) # convert gene names to Entrez ID genemap <- select(org.Mm.eg.db, selected, "ENTREZID", "SYMBOL") univmap <- select(org.Mm.eg.db, universe, "ENTREZID", "SYMBOL") # load GO scoring package biocLite(“GOstats”) library(GOstats) # set up analysis param<- new ("GOHyperGParams", geneIds = genemap, universeGeneIds=univmap, annotation="org.Mm.eg.db", ontology="BP",pvalueCutoff=0.01, conditional=FALSE, testDirection="over") # run analysis hyp<-hyperGTest(param) # visualize summary(hyp) ## Select/sort on Pvalue, Count, etc.
- 36. Summary • Intro of RNA-‐seq • EsCmaCng expression levels • DifferenCal expression analysis with DESeq2 • Andrea Sboner: ans2077@med.cornell.edu