TL;DR - I ran Tophat2 on human RNA-seq data and the mapping rate was only 9%. How can I determine if my sequencing data is junk, i.e. something went wrong with the sequencing.
Total beginner to RNA-seq analysis here.
I received 4 RNA profiling sequencing data, 3 for mouse and 1 for human. I ran Fastqc on all of them, and then I ran a trimmer code (which we had designed earlier and does the following - remove the first 3 bases, the last 7 bases, discard any read if it is smaller than 15 based etc..). Then I ran tophat on all the samples, with the only modification that the gtf file and the fasta file only has information about exons (this was because earlier Bowtie used to create problems with the full files). The tophat code -
tophat2 -o ./tophat/$x --transcriptome-index ~/.../mod.gencode.v19 --no-coverage-search --keep-tmp --num-threads 14 ~/...mod.GRCh37.p13.genome ./fastq/human/$x.fastq
(Of course, I used mouse gtf,fasta files for the mouse samples and the human files for human samples).
For all the 3 mouse samples, the mapping rate was around 50%, which is what we have seen for our samples before too. For the human sample, it was only 9%.
Now I want to know if the problem is with the sequencing or somehow I can do something to come up with a better mapping. A friend said that I should check the levels of some highly expressed genes (using the mapped bam file) on IGV, and see if the levels are what you would expect. So, I wanted to ask the community, as to what I can do to test if the data is indeed faulty, or if the mapping can be improved.
Note - it is sequenced by a Next Gen sequencer
Here is the Fastqc file for the human sample - https://drive.google.com/file/d/1hzotX-G2Bn0-SKmI8oX_pUNxokh76iT3/view?usp=sharing
With just 9% alignment you can pretty much take any sample of the data (do not use the reads at top of the file) and blast them at NCBI as has been suggested.
Has this data been trimmed? If not try that first. There may be a big 3' bias (poly-A's in your data as well).
Try a different aligner (I suggest
bbmapto see what you get. If the data is junk then no miracles are likely but you will at least convince yourself of that fact.Is this sequenced on a NextSeq?
Yes it is on a next gen sequencing machine
The overrepresented sequences suggest this is from a Illumina NextSeq, with the polyG reads. Alternatively another sequencer using two colour labels.
Have you tried taking some of the unaligning reads and putting them into NCBI blastn? that will tell you what these are, I'm guessing either contamination from another species or unremoved rRNA
Thanks Philip for your comment. How can I get these unaligned reads? I know that tophat creates 2 bam files - one for mapped hits, and another for the unmapped, but I don't know a way to extract the unaligned sequences from the bam file as it's a binary file. Any ideas?
If you run
samtools viewon your bam file of unaligned reads you'll get the reads in SAM format, from there you'll have to copy paste a few into the search window of NCBI blastnHave you tried aligning the human reads to mouse, just to check if something went wrong there?
The data look quite low quality to me...
So, do you mean that I should try to map the human fastq files to the mouse genome using the mouse gtf and fasta files? If so, is your hunch that the data is coming mostly from mouse sample somehow?
I don't know - just need to exclude potential contamination. Although (haven't tried) I would expect more human reads to map to the mouse genome due to homology.