RNA-Seq: Low mapping percentage of pair-end reads (length 75bp)

[wilson90]

Dear community,

I have this Mus Musculus pair-end RNA-Seq disease sample of length 75bp. I mapped it onto Mus Musculus genome (NCBI Build 37.2) using Tophat, and I have found that the percentage of reads with at least one mapping is extremely low, 25%.

On the other hand, the control sample has 80% mapping percentage.

I have checked for possible contamination by mapping the samples with human genome, mycoplasma genome, and E.Coli genome, and I am pretty sure that possibility of contamination can be ruled out because the percentage mapping to these genomes are less than 1%.

My question: What are the common causes for extremely low percentage of mapping? Errornous base-calling? High error rate in sequencing steps? or Tophat is not good enough? Or can I conclude this as a new finding with further validation?

Tell me if you need more information. =)

[pmcget]

Do you have the same read lengths for the control sample - or are they shorter?

[pmcget]

If you want to search a wider range for potential contaminants you could use MEGAN:

http://ab.inf.uni-tuebingen.de/software/megan/

[wilson90]

Dear pmcget,

The read lengths are the same. Both 75bp. How I wish that this is due to biological events rather than contamination or sequencing errors. =D

I am currently checking the reads using FASTQC.

Thank you for the suggestion. I will definitely try that out and let you know if it helps. =)

[wilson90]

Just a quick question:

Does presence of poly-A tail in the reads affect mapping? Because base-level analysis does shown higher percentage of A bases at the end of my reads.

[pmcget]

Differing amounts of PolyA almost certainly will lead to mapping bias.

A quick search found this tool for trimming polyA from FASTQ files:

http://chipster.csc.fi/manual/prinseq-AT-trimmer.html

You could trim both files and compare results after trimming.
Differing polyA could be library prep/sequencing artifact or biological.

Do you have multiple replicates? If so do they all show same effect?
Might help to exclude potential explanations

[alexdobin]

Just a quick question:

Does presence of poly-A tail in the reads affect mapping? Because base-level analysis does shown higher percentage of A bases at the end of my reads.

You can try to map with STAR without any pre-trimming. STAR will trim as many bases from the ends as needed to map each read. Checking the bases that were trimmed by STAR may help you to figure out what happened with your tails.