1. RNA-Seq Visualization
cummrRbund in Atmosphere
Jason Williams
iPlant / Cold Spring Harbor Laboratory

2. *Graphics taken from these publications

3. The Tuxedo Protocol
*TopHat and Cufflinks require a sequenced genome

4. Tophat Explain reference-sequence based NGS read alignments.
Explain that we are skipping the cufflinks step because the Arabidopsis transcriptome is so well annotated that we can use the TAIR gene models as our refernce transcripts for CuffDiff

5. TopHat outputs in IGV

6. Using CummeRbund in Atmosphere
Explain that we are skipping the cufflinks step because the Arabidopsis transcriptome is so well annotated that we can use the TAIR gene models as our refernce transcripts for CuffDiff

7. Using CummeRbund in Atmosphere
Visualize and mine Cuffdiff results
Output files from Cuffdiff are reorganized into a local database
Explain that we are skipping the cufflinks step because the Arabidopsis transcriptome is so well annotated that we can use the TAIR gene models as our refernce transcripts for CuffDiff

8. Choose the right image We will be using “RNA-Seq Visualization”
Rmi-BE9C2D12
Any image w/R can work, and you could also search
For an image with cummeRbund installed
Explain that we are skipping the cufflinks step because the Arabidopsis transcriptome is so well annotated that we can use the TAIR gene models as our refernce transcripts for CuffDiff

9. Installing cummeRbund in R

10. Installing cummeRbund in R

11. Reading the data in > cuff <- readCufflinks() > cuff
CuffSet instance with:
2 samples
33714 genes
43481 isoforms
35113 TSS
32924 CDS
33621 promoters
35113 splicing
27350 relCDS

12. Visualizing dispersion
>disp<-dispersionPlot(genes(cuff))
>disp
Counts vs. dispersion
Overdispersion greater variability in a data set than would be expected based on a given model ( in our case extra-Poisson variation)
If you use Poisson model, you will overestimate differential expression

13. Visualizing dispersion
Poisson adequately describes technical variation

14. Visualizing dispersion

15. Squared-coefficient of Variation (SCV)
>genes.scv<-fpkmSCVPlot(genes(cuff))
>genes.scv
Normalized measure of cross-replicate variability
Represents the relationship of the standard deviation to the mean
Differences in SCV can result in lower numbers of differentially expressed genes due to a higher degree of variability between replicate fpkm estimates

16. Distributions of FPKM scores across samples
>dens<-csDensity(genes(cuff))
>dens
>densRep<-csDensity(genes(cuff),replicates=T)
>densRep
Non-parametric estimate of pdf

17. FPKM Pairwise Scatter Plots
> csScatter(genes(cuff),‘WT’,‘hy5’,smooth=T)

18. Saving your Plots
1. Plot type: >(e.g. jpeg, png, pdf) (file_path_and_file_name)
2. Plot function
3. dev.off()
> png (‘csScatter.png’) #Will save in working directory
> csScatter(genes(cuff),‘WT’,‘hy5’,smooth=T)
>dev.off

19. Selecting and Filtering Gene Sets
Using the ‘getSig’ function
# Enables you to get genes at significance n
>sig <-getSig(cuff, alpha=0.05, level =‘genes’) # genes of significance 0.05
>length(sig) #returns the number of genes in the sig object
>sig <-getSig(cuff, alpha=0, level=‘genes’)
>tail(sig,100) #displays the last 100 genes in the sig object you just made

20. Selecting and Filtering Gene Sets
Using the ‘getGenes’ function
# Get the gene information
>sigGenes <- getGenes(cuff,sig)
Plot this in another scatter plot
>csScatter(sigGenes, ‘WT’, ‘hy5’)

21. Heat mapping Similar Expression Values
>sigGenes <-getGenes(cuff,tail(sig,50))
#last 50 genes in the list we created of genes
>csHeatmap(sigGenes,cluster=‘both’)

22. Heat mapping Similar Expression Values
>csHeatmap(sigGenes,cluster=‘both’,replicates=‘T’)

23. Expression Plots by Genes
> myGeneId<-”AT5G41471"
> myGene<-getGene(cuff,myGeneId)
> myGene

24. Expression Plots by Genes
> expressionPlot(myGene,replicates=‘T’)