1. Introduction to RNAseq

2. NGS - Quick Recap
Many applications -> research intent determines technology platform choice
High volume data BUT error prone
FASTQ is accepted format standard
Must assess quality scores before proceeding
‘Bad’ data can be rescued

3. The Central Dogma of Molecular Biology
Reverse
Transcription

4. RNAseq Protocols cDNA, not RNA sequencing
Types of libraries available:
Total RNA sequencing (not advised)
polyA+ RNA sequencing
Small RNA sequencing (specific size range targeted)

5. cDNA Synthesis

6. Genome-scale Applications
Transcriptome analysis
Identifying new transcribed regions
Expression profiling
Resequencing to find genetic polymorphisms:
SNPs, micro-indels
CNVs
Question: Why even bother with exome sequencing then?

7. What about microarrays??!!!
Assumes we know all transcribed regions and that spliceforms are not important
Cannot find anything novel
BUT may be the best choice depending on QUESTION

8. Arrays vs RNAseq (1)
Correlation of fold change between arrays and RNAseq is similar to correlation between array platforms (0.73)
Technical replicates almost identical
Extra analysis: prediction of alternative splicing, SNPs
Low- and high-expressed genes do not match

9. RNA-Seq promises/pitfalls
can reveal in a single assay:
new genes
splice variants
quantify genome-wide gene expression
BUT
Data is voluminous and complex
Need scalable, fast and mathematically principled analysis software and LOTS of computing resources

10. Experimental considerations
Comparative conditions must make biological sense
Biological replicates are always better than technical ones
Aim for at least 3 replicates per condition
ISOLATE the target mRNA species you are after

11. Analysis strategies De novo assembly of transcripts:
+ re-constructs actual spliced transcripts
+ does not require genome sequence
easier to work post-transcriptional modifications
- requires huge computational resources (RAM)
- low sensitivity: hard to capture low abundance transcripts
Alignment to the genome => Transcript assembly
+ computationally feasible
+ high sensitivity
+ easier to annotate using genomic annotations
- need to take special care of splice junctions

12. Basic analysis flowchart
Illumina
reads
Remove
artifacts
AAA..., ...N...
Clip adapters
(small RNA)
"Collapse"
identical
reads
Align
to the
genome
Pre-filter:
low complexity
synthetic
Count
and
discard
Re-align
with different
number of mismatches
etc
un-mapped
mapped
mapped
un-mapped
Assemble:
contigs (exons)
+ connectivity
Filter out low
confidence
contigs
(singletons)
Annotate

13. Software Short-read aligners Data preprocessing Expression studies
BWA, Novoalign, Bowtie, TOPHAT (eukaryotes)
Data preprocessing
Fastx toolkit, samtools
Expression studies
Cufflinks package, R packages (DESeq, edgeR, more…)
Alternative splicing
Cufflinks, Augustus

14. Very widely adopted suite
The ‘Tuxedo’ protocol
TOPHAT + CUFFLINKS
TopHat aligns reads to genome and discovers splice sites
Cufflinks predicts transcripts present in dataset
Cuffdiff identifies differential expression
Very widely adopted suite

16. Read alignment with TopHat
Uses BOWTIE aligner to align reads to genome
BOWTIE cannot deal with large gaps (introns)
Tophat segments reads that remain unaligned
Smaller segments mostly end up aligning

17. Read alignment with TopHat (2)

18. Read alignment with TopHat (3)
When there is a large gap between segments of same read -> probable INTRON
Tophat uses this to build an index of probable splice sites
Allows accurate measurement of spliceform expression

19. Cufflinks package http://cufflinks.cbcb.umd.edu/ Cufflinks: Cuffdiff:
Expression values calculation
Transcripts de novo assembly
Cuffdiff:
Differential expression analysis

20. Cufflinks: Transcript assembly
Assembles individual transcripts based on aligned reads
Infers likely spliceforms of each gene
Quantifies expression level of each

21. Cuffmerge Merges transfrags into transcripts where appropriate
Also performs a reference based assembly of transcripts using known transcripts
Produces single annotation file which aids downstream analysis

22. Cuffdiff: Differential expression
Calculates expression level in two or more samples
Expression level relates to read abundance
Because of bias sources, cuffdiff tries to model the variance in its significance calculation

23. FPKM (RPKM): Expression Values
Fragments Reads Per Kilobase of exon model per Million mapped fragments
Nat Methods. 2008, Mapping and quantifying mammalian transcriptomes by RNA-Seq. Mortazavi A et al.
C= the number of reads mapped onto the gene's exons
N= total number of reads in the experiment
L= the sum of the exons in base pairs.

24. Cuffdiff (differential expression)
Pairwise or time series comparison
Normal distribution of read counts
Fisher’s test
test_id gene locus sample_1 sample_2 status value_1 value_2 ln(fold_change) test_stat p_value significant
ENSG TSPAN6 chrX: q1 q2 NOTEST no
ENSG TNMD chrX: q1 q2 NOTEST no
ENSG DPM1 chr20: q1 q2 NOTEST no
ENSG SCYL3 chr1: q1 q2 OK yes

25. Recommendations You can use BOWTIE or BOWTIE2 but Use CUFFDIFF2
Better statistical model
Detection of truly differentially expressed genes
VERY easy to parse output file (See example on course page)