Introduction to Transcriptomics

Contributors

Authors:

Bérénice Batut

Anika Erxleben

Markus Wolfien

last_modification Last modification: Dec 20, 2021

What is RNA sequencing?

RNA

The structure of a eukaryotic protein-coding gene

Transcribed form of the DNA
Active state of the DNA

.footnote[Credit: Thomas Shafee]

RNA sequencing

Summary of RNA-Seq principle. In vivo transcription, pre-mRNA, intron splicing all rpoduce a mature mRNA. In vitro this is fragmented into RNA fragments, reverse transcribed into double stranded cDNA and then sequenced.

RNA quantification at single base resolution
Cost efficient analysis of the whole transcriptome in a high-throughput manner

.footnote[Credit: Thomas Shafee (adapted)]

Where does my data come from?

select a cell population and extract total RNA is shown at the top. Small RNA are size selected by PAGE or kit, an adapter ligated, and converted to cDNA. Or poly(a) selects ribosome minus, and those mRNA are fragmented, and converted to cDNA. In both cases the cDNA becomes a library for sequencing.

.footnote[Zang and Mortazavi, Nature, 2012]

Principle of RNA sequencing

Cartoon showing a magazine stand labelled transcriptome, and a person saying "I'll take all of them". These are run through a shredder, before hundreds of people attempt to re-assemble, and the person hands the professor a poorly assembled magazine.

.footnote[Korf, Nat Met, 2013]

Challenges of RNA sequencing

Different origin for the sample RNA and the reference genome
Presence of incompletely processed RNAs or transcriptional background noise
Sequencing biases (e.g. PCR library preparation)

Benefits of RNA sequencing

A word cloud highlighting words like benefits, novel, sensitivity.

2 main research applications for RNA-Seq

Transcript discovery

Which RNA molecules are in my sample?

Novel isoforms and alternative splicing, Non-coding RNAs, Single nucleotide variations, Fusion genes
RNA quantification

What is the concentration of RNAs?

Absolute gene expression (within sample), Differential expression (between biological samples)

How to analyze RNA seq data for RNA quantification?

RNA quantification

$Select RNA fraction of interest (polyA, ribo-minus, and others), these are fragmented and reverse transcribed before sequencing and mapping onto the genome and quantification.$

.footnote[Pepke et al, Nat Met, 2009]

Overview of the Data Processing

Control and treatment files goes through QC, annotation, and rad counting to produce sets of count tables. Then differential expression analysis is computed.

No available standardized workflow
Multiple possible best practices for every dataset

Data Pre-processing

Adapter clipping to trim the sequencing adapters
Quality trimming to remove wrongly called and low quality bases

.footnote[See NGS Quality control]

Annotation of RNA-Seq reads

Simple mapping on a reference genome? More challenging

A cartoon of a pre-mRNA, intro and exons. These map to an mRNA and short reads are shown piled up against the mRNA. The short read is splity by intron when aligning to a reference genome.

.footnote[Credit: Rgocs]

Annotation of RNA-Seq reads

3 main strategies for annotations

Transcriptome mapping
Genome mapping
De novo transcriptome assembly and annotation

Transcriptome mapping

Cartoon of multiple exons collapsed, and paired end reads being shown as easy to align.

See NGS Mapping

Need reliable gene models
No detection of novel genes

.footnote[Figures by Ernest Turro, EMBO Practical Course on Analysis of HTS Data, 2012]

Genome mapping

Splice-aware read alignment

The same cartoon again, but now it is shown split up by introns, and one of the paired end reads is split across three exons, so it is hard to align.

Detection of novel genes and isoforms

.footnote[Figures by Ernest Turro, EMBO Practical Course on Analysis of HTS Data, 2012]

Transcriptome and Genome mapping

Needed

Reference genome/transcriptome in FASTA
Annotations of known genes, … in GTF

Where to find?

Joint projects to produce and maintain annotations on selected organisms: EMBL-EBI, UCSC, RefSeq, Ensembl, …

De novo transcriptome assembly

No need for a reference genome …

Assembly into transcripts
Map reads back

Quantification

What is the expression level of the genomic features?

Counting the number of reads per features: Easy!!
Challenges
- How to handle multi-mapped reads (i.e. reads with multiple alignments)?
- How to distinguish between different isoforms?
  - At gene level?
  - At transcript level?
  - At exon level?

Differential Expression Analysis

.image-75[ Three conditions are created and multiple transcriptomics sequenced into reads and mapped and compared. ]

Account for variability of expression across biological replicates
with the help of counts

Differential Expression Analysis: Normalization

Make the expression levels comparable across

By Features: genes, isoforms
By Samples
Methods
- FPKM/RPKM (Cufflinks/Cuffdiff)
- TMM (edgeR)
- DESeq2 (DESeq2)

.footnote[“Only the DESeq and TMM normalization methods are robust to the presence of different library sizes and widely different library compositions…” - Dillies et al., Brief Bioinf, 2013]

Impact of sequencing depth and number of replicates

.image-75[ Image of a table from a paper. The recommendation is at least three biological replicates to accurately detect changes. 3 replicates will give you an 87% chance of detecting a 2-fold change, but only a 17% chance of detecting a 1.25 fold change. ]

.footnote[Conesa et al, Genome Biol, 2016]

Recommendation: At least 3 biological replicates

Speaker Notes

Number of replicates has greater effect on DE detection accuracy than sequencing depth (more replicates = increased statistical power)
DE detection of lowly expressed genes is very sensitive to number of reads and replication
DE detection of highly expressed genes possible already at low sequencing depth

Visualization

Integrative Genomics Viewer (IGV) or Trackster

Visualization of the aligned BAM files
Sashimi plots

Quantitative visualization of read coverage along exons and splice junctions
CummeRbund

Visualization package for Cufflinks high-throughput sequencing data

Thank you!

This material is the result of a collaborative work. Thanks to the Galaxy Training Network and all the contributors!

This material is licensed under the Creative Commons Attribution 4.0 International License.