UAB Galaxy RNA Seq Step by Step Tutorial
Transcriptome analysis via RNA-Seq
There are several types of RNA-Seq: transcriptome, splice-variant/TSS/UTR analysis, microRNA-Seq, etc. This tutorial will focus on doing a 2 condition, 1 replicate transcriptome analysis in mouse.
- Jeremy Goecks' Galaxy RNAseq tutorial http://main.g2.bx.psu.edu/u/jeremy/p/galaxy-rna-seq-analysis-exercise
- Tophat short-read RNA mapper http://tophat.cbcb.umd.edu/
- Cufflinks transcript assembler http://cufflinks.cbcb.umd.edu/
- Tophat-cuff[links,compare,diff] tutorial (command-line) http://cufflinks.cbcb.umd.edu/tutorial.html
- IGV (Broad Integrated Genome Viewer) http://www.broadinstitute.org/software/igv/download
- IGB (Integrated Genome Browser) http://bioviz.org/igb/download.shtml
- UCSC Genome Browser http://genome.ucsc.edu/
File Types and Acronyms
- FASTQ - text file; like a FASTA but with extra lines for read quality scores for each base. This is what you usually get from the sequencing center.
- BAM - Binary compresssed sequences Alignment/Map format - Contains a copy of the reference genome, and every short read that was mapped, including the short reads location and alignment. It can also contain the "left-over" reads from the source FASTQ's that did NOT align to the reference genome! These files must visualized with a tool, either IGV, IGB or sometimes UCSC Genome Browser.
- BED - text file; a list of locations, in genomic coordinates.
- GTF Gene Transfer Format - text file; contains gene/transcript annotations for a sequence/genome, also in genomic coordinates.
- FASTA - text file; contains a list of named sequences. Can be used to hold the sequence of a reference genome to which RNAseq reads will be mapped.
- FPKM (RPKM) - acronym; Fragments Per Kilobase of exon per Million fragments mapped
For this tutorial, we will have 2 mouse samples, sequenced with paired-end reads on an Illumina machine. That gives us 4 FASTQ files to upload (forward and reverse sequences for each sample).
FTP site for compressed fastq.gz files
Or they can be found inside Galaxy at Shared Data / Data Libraries / Tutorial Data Sets / RNAseq Tutorial
Set filetype and genome
To start, you must move the data (FASTQ) from the sequencing center into the Galaxy instance, be sure to specify the filetype ("fastqsanger" for UAB and HudsonAlpha, the default "fastq" will not work) and the organism that was sequenced ("Genome database"), in this tutorial, "mm9" for mouse.
Check quality of READ data
At this step, we check the quality of sequencing. This was already covered in the DNA-Seq tutorial, under Galaxy_DNA-Seq_Tutorial#Assessing_the_quality_of_the_data Assessing the quality of the data. This says nothing about the quality of the sample, or whether it was the right sample. We'll check that later.
At a minimum, for each FASTQ file, run NGS: QC and manipulation > FASTQ QC > Fastqc
This quick tool will give you a nice HTML report, including dashboard summary. If some of the reports are not green, discuss this with your sequencing center.
In some cases, you may need to trim bases off the beginning or ends of the sequence. See the tools in NGS: QC and manipulation > FASTX-TOOLKIT FOR FASTQ DATA.
Align using TopHat: inputs
For the moment, TopHat is the standard NGS aligner for transcript data, as it handles splicing. In addition, it can be set to detect indels relative to the reference genome.
We'll run TopHat once for each sample (twice, in this case), providing it with 2 FASTQ files each time (forward and reverse reads).
NGS: RNA Analysis > Tophat for Illumina
- Reads (FASTQs)
- FASTQs As we're doing "paired" reads, we will need to provide 2 FASTQ files: the forward and reverse reads. In order to have a place to specify the 2nd FASTQ file, we must set the "Is this library mate-paired?" pulldown to "Paired-end", then a second pulldown will appear to specify the 2nd FASTQ.
- Mean Inner Distance between Mate Pairs is a value you must get from your sequencing center. This is the mean fragment length of the molecules sequenced, minus the part sequenced. For many RNAseq experiments, it is around 150-175.
- Genome (Built-in or FASTA from history)
- We must provide it with the reference genome to align to. In our case, this is Mouse, and we'll use the already installed "mm9" genome build. If you have an genome that is not on the list, you can either have us add it, or you can upload a FASTA file of the genome into your history, and point TopHat at that.
- TopHat settings to use
- This is the easy thing to do, but, as the manual said, "There is no such thing (yet) as an automated gearshift in splice junction identification. It is all like stick-shift driving in San Francisco. In other words, running this tool with default parameters will probably not give you meaningful results."
- Full parameter list
- The most common thing to change
- Allow indel search (from NO to YES)
- Minimum isoform fraction (to 0 if looking for rare isoforms)
- Maximum/minimum intron length (defaults are for Mammals; other critters will do better with stricter settings)
- The most common thing to change
- accepted_hits (BAM, BAI)
- Two binary files: .BAM (data) and .BAI (index)
- These are the actual paired reads mapped to their position on the genome, and split across exon junctions. This can be visualized in IGV or IGB, but you must download both .BAM and .BAI files to the same directory.
- splice_junctions (BED)
- BED file (list of genomic locations, no sequence) listing all the places TopHat had to split a read into two pieces to span an exon junction. This can be visualized at UCSC or in IGV, etc.
- deletions (BED) (if indel search is on)
- insertions (BED) (if indel search is on)
Next we check the quality of the mapping to the reference genome. This will detect not only problems with the sequencing runs, but also contamination or swapping of the samples.
For each sample, run NGS: SAM Tools > flagstat on the accepted_hits.bam file from tophat. While the same tool is used to assess the output of other short-read genomic aligners (BWA, bowtie, etc), the values are interpreted a little differently.
# flagstat output 28316 in total 0 QC failure 0 duplicates 28316 mapped (100.00%) 28316 paired in sequencing 14258 read1 14058 read2 16398 properly paired (57.91%) 28010 with itself and mate mapped 306 singletons (1.08%) 0 with mate mapped to a different chr 0 with mate mapped to a different chr (mapQ>=5)
Things to ignore
- in total/mapped (100.00%): these numbers always match and give 100% for tophat runs.
- The in total can also be greater than the number of reads in the original FASTQ, as tophat can create multiple matches for a read.
Things to look at
- properly paired should be above 50%, the higher the better. This is the count of read-pairs (sequences from the opposite ends of the same physical molecule) that mapped within the expected distance of each other, and in the expected orientation. Ideally this should be 100%.
- If properly paired is low the mapping quality may be bad, or there may be sample contamination.
Cufflinks takes the mapped reads for each sample and tries to reconstruct the individual transcripts and isoforms. This can be done "de novo", without using any previous gene annotations, limited strictly to only computing expression levels of existing annotations, or a combination of the two.
For this exercise, we are comparing transcription levels between samples for known genes, so we will just compute expression of known transcripts.
* UAB Modified Reference annotations