RNA-seq presentation

Please DO NOT switch on your
computers – yet.
RNA-seq Analysis
Graham Etherington
Sainsbury Laboratory Training Course
Today's topics
• The basics – What is RNA-seq, paired-end reads, alternative
• Considerations before sequencing
– Library prep
– What ‘contaminate’ RNA (rRNA, abundant transcripts) to
remove and how.
• Sequencing
– Quality control
– Assembly techniques
• Reference-based alignment
• De-novo assembly
• Combined assembly (Align-then-assemble vs Assemble-then-align)
• Choosing a strategy and a program
• Expression analysis
Today's topics
• Tutorials
– Reference-based transcript assembly
and expression analysis without
annotation using Galaxy
• TopHat – Cufflinks - Cuffmerge - Cuffdiff
– De-novo assembly using Trinity
What is RNA-seq?
Extract mRNA (expressed genes)
Sequence mRNA
Assemble into
RNA-seq basics - Paired-end reads
Sequences can be paired-end
– sequences occur as ‘pairs’ with one left-hand (forward) read and one right-hand
(reverse) read.
– a given distance (insert-size) between the start and end of pairs.
Paired -ends
Left (forward) read
76 nucleotides
Right (reverse) read
76 nucleotides
500 nt DNA fragment
~350 nt gap
~500 nt ‘insert size’
RNA-seq basics - Alternative splicing
RNA-seq – the basics
Genome of interest.
How many genes (mRNAs) are there?
Are some novel?
Alternative spliced isoforms?
Which genes are expressed under different
environmental conditions (cf microarrays)?
• Are some expressed more than others?
• Library prep.
• Multiple insert sizes captures both short and long
transcripts plus alternative spliced isoforms
– longer insert sizes offer long-range exon
• Which RNA to select
– poly-A tail RNA
– misses ncRNA + rare mRNAs without poly-A tail
– leave all RNAs in then remove rRNA by
‘hybridisation-based depletion methods’
• biases quantification of high-abundant transcripts
• Strand-specific protocols
– Aids assembly and quantification of overlapping
transcripts from opposite strands
• Quality control
• LOTS of data – don’t worry
about throwing a lot of it away
– remove short/long reads
– remove reads with Ns
– remove PCR duplicates
– remove/trim low-quality
– Remove low copy k-mers
Reference-based Alignment
• Use when a closely-related reference is
• 3 steps
① Use a splice-aware aligner (e.g. BLAT, TopHat).
② Cluster reads from each locus to build isoform
③ Traverse graph to resolve isoforms (e.g.
Cufflinks, Scripture)
Splice-aware aligners
• Two types- Seed & extend and BWT
SEED-part of read
EXTEND alignment
Splice-aware aligners
• Burrow-Wheeler transform (BWT)
• Creates a compressed ‘index’ of the genome.
• Stretches of sequence can be ‘looked-up’
– Narrows-down the search space
– Speeds up alignment
– Requires less memory
Creating and Traversing Graphs
Reference-based Alignment
• Applications:
– Microbes and lower eukaryotic organisms.
– Few introns and little alternative splicing
– Use with strand-specific sequencing to identify
overlapping genes.
Reference-based Alignment
• Advantages:
– Contamination not a great problem – won’t align.
– Less memory use
– Align low-abundance transcripts
– Identify transcripts undiscovered in annotated
Reference-based Alignment
• Disadvantages:
– Relies on the accuracy of the reference sequence
• May contain errors, deletions, missassemblies.
• Can miss divergent transcripts
– Reads often align to multiple regions
• Excluding multi-mapped reads – leaves gaps
• Randomly assign multi-mapped reads – false transcripts
– Can’t easily assemble trans-spliced genes
Reference-based Alignment
• Summary
• Preferable where a high-quality reference
• Can assemble full-length transcripts at depth
of 10x.
• Can include longer reads (e.g . 454) to capture
connectivity between more exons.
De-novo assembly
• Doesn’t use a reference sequence.
• Finds overlaps between reads and assembles
them into contigs/transcripts.
• Constructs De Bruijn graph which breaks reads
into k-mers and connects overlapping nodes.
De Bruijn graphs
All substrings of length k (k-mers) are generated from each read.
De Bruijn graph created by kmers that overlap by k–1.
Single-nucleotide differences cause 'bubbles' of length k in the
De Brujin graph
Insertions or deletions introduce a shorter path in the graph.
Collapse adjacent nodes.
Calculate paths through graph.
De-novo Assembly
• Applications:
– Microbes and lower eukaryotic organisms.
– Yeast transcriptomes can be assembled with >30x
– Overlapping genes from opposite strands can be
detected by not allowing reverse complements in
De Bruijn graph and using odd k-mers.
– Higher eukaryotes more challenging due to larger
datasets and difficulties in identifying alternative
splice sites.
De-novo Assembly
• Advantages
– Doesn’t need a reference sequence.
– Sometimes better than reference-based assembly
• reference is of low quality (e.g. missing bits).
• Unknown exogenous transcripts want to be detected.
• Where long introns are expected.
– Doesn’t depend on the correct alignment of reads
to splice sites.
De-novo Assembly
• Disadvantages:
– With higher eukaryotic datasets needs lots of
– Requires higher sequencing depth than referencebased assembly (30x cf 10x).
– Highly similar transcripts are likely to be
assembled into single transcripts.
– Sensitive to read-errors. Hard to tell errors from
low-abundance transcripts.
Combined strategy
• Use both de-novo assembly and referencebased alignment methods to get the best
• Two techniques:
– Align-then-assemble
– Assemble-then-align
• Make use of sensitivity of reference-based
aligners and use de-novo assembly for novel
Combined strategy
• Align-then-assemble
– Most intuitive.
– Align reads to a reference.
– What doesn’t align – denovo assemble.
Combined strategy
• Assemble-then-align
– When quality of reference
genome is suspect.
– When reference genome is from
distantly related species.
– De-novo assemble into contigs
– Then use reference to extend
contigs into longer transcripts.
– Small errors in the reference
genome don’t get propagated
into the new assembly.
Choosing a strategy
• Factors to consider
– Reference genome available?
• Good quality?
• Closely-related species?
– Aim of project
• Annotation
• Identify novel transcripts
• Expression analysis
Choosing a splice-aware alignment
Choosing a transcript assembly
Expression analysis
The more abundant an RNA, the more times it will be randomly selected for sequencing.
Gene 1
Condition A
Gene 1
Condition B
expressed mRNA
Expression analysis
• Use No. of mapped reads as an indicator of
Map reads back to genome
Gene 1
Condition A
Gene 1
Condition B
Expression analysis
• Need some way to normalise the expression
• Fragments Per Kilobase of exon per Million
fragments mapped (FPKM).
• Some controversy over this approach – bias
for longer transcripts.
Switch on your computers and boot into Windows.
Log-in using the yellow username on your machine.
Go through the tutorial sheet.
There are two tasks, both using Galaxy:
– Reference-based transcript assembly and expression analysis
without annotation using Galaxy
• TopHat – Cufflinks - Cuffmerge - Cuffdiff
– De-novo transcript assembly using Trinity.
• Take your time during the tutorials and make sure you
understand what you are doing.
• Please delete your Galaxy analysis when finished.
• Logging on to your computers:
– Use the name given on the yellow sticker on your
– Password: Learning26
• Logging into Galaxy
– Go to http://galaxy.tsl.ac.uk
– [email protected] (e.g. [email protected])
– Password: Learning26

similar documents