A Question About Pair-End Reads Alignment

Question

[Tonyzeng]

hi, I have a simple question, When I align pair-end reads with reference, if my reverse reads are bad reads which can not be modified even I did QC, can I just use forward reads for alignment without reverse reads? What situation that I can give up use reverse or forward reads?

Answer 1

[Biomonika (Noolean)]

Please try to avoid stating that the question you are asking is simple :-)

If reverse reads are of bad quality, multiple strategies can be employed, for example:

1, if one read from the pair is of the bad quality, whole pair may be excluded from the further analysis so that you keep only read pairs with high quality base calls

2, you may want to keep only one read. This will raise the problem of having two types of data: paired-end reads and single reads and therefore the need of two rounds of mapping (mapping software can do typically either paired-end mapping or single)

I believe that removing the second read will do nothing but decrease success of your mapping - pairs help to determine the right position in the genome more precisely. Therefore, you might get lower mapping score or even misalign the reads to the wrong position.

Answer 2

[Ian]

If your forward reads are "good" and the reverse are "bad", perform fragment mapping with your forward reads. The effect also depends on the type of analysis. If you are just counting, e.g. ChIP-seq, you'll be OK. If you are performing RNA-seq analysis and need to know the splice variants, then you have a problem.

I know you have run QC, but I can recommend Trimmomatic as it allows QC filtering of paired reads and maintains the order of filtered full pairs in the output.