% of reads mapped to transcriptome

[cement_head]

Hello,

I am doing RNA-Seq analysis and needed some advice on mapping PE reads back to the reference transcriptome. I made the reference from the set of 12 RNA-Seq reads (after duplicates) were removed.

I am then mapping each RNA-Seq dataset back to the transcriptome, but I am only getting about 60% mapping back to the reference transcriptome. Is it possible that my fragment length is too short, thus precluding reads to be mapped back to my reference? I have the mean size of the fragments from the sequencing facility and I've set the program (CLC GW) to accomodate a range 75 bp below and above the mean size (fragment length between 150 and 300).

Or should I start looking to relax some of the parameters that influence the matching of reads to the reference?

Thanks,
Andor

[peromhc]

Andor,

You should consider the quality of your transcriptome assembly as, in my experience, CLC will very often do a poor job of reconstruction. For instance, how many full length transcripts do you have? How many match to transcripts from the closest reference using a reciprocal best blast procedure?

[nucacidhunter]

Your reference is basically singletons or contigs from assembly of non-redundant reads from all libraries. So, all of them should map back to the reference.

[cement_head]

Andor,

You should consider the quality of your transcriptome assembly as, in my experience, CLC will very often do a poor job of reconstruction. For instance, how many full length transcripts do you have? How many match to transcripts from the closest reference using a reciprocal best blast procedure?

So I did an assembly on CLC using 454 data and it was very good - 85% got a blastx hit. But this illumina data, not so much. Might try a different assembler...

[cement_head]

Your reference is basically singletons or contigs from assembly of non-redundant reads from all libraries. So, all of them should map back to the reference.

right...theoretically...or at least a VERY high percentage...