how to utilize unmapped paired-end data

[ZoeG]

Hi, guys, I am thinking the way to utilize those unmapped reads.
When I mapped PE100 data using TopHat, some reads were threw out as unmapped reads.
When I looked into these unmapped reads, though a lot are garbage or mis-matched reads, I found some of them contained information. Thus, I am thinking to utilize these unmapped data. I gave TopHat a specific GTF and asked the data mapped to this GTF with -G option.
However, I found that though with a good mapping rate (>80%), only a small percentage of reads are properly aligned (~20%). I think that's because the unmapped data of paired-end reads ( left vs right) are not synchronized.
Can I synchronize the unmapped left and right reads with some recommended tools?
Or do you think if this idea worth time spending?
Thanks,

[Cofactor Genomics]

Hi ZoeG,
I wonder what your ultimate goal is in the project.

My team uses these unmapped reads many times in a way to build confidence in a possible insertion sequence between the reference and the sequenced organism by identifying one leg of the pair mapping to the reference and another that does not. This information in concert with other key coverage characteristics provides these insertion results.

However, if you are ultimately after SNPs and small INDELs the inclusion of this data is not worth it... it only casts doubt on your more confident mappings.

My best.

Jarret Glasscock
Cofactor Genomics

http://www.cofactorgenomics.com

[ZoeG]

Thanks for the advice, Jarret.
I am thinking whether it is possible to map the long genes first and then, map relatively small genes in the unmapped reads. I hope that the mis-matches for short genes could be decreased in this way.
But there may be a risk of a loss of reads containing short genes ?

How to align the unmapped reads is also a question for me since I am relatively new to the field.

Hi ZoeG,
I wonder what your ultimate goal is in the project.

My team uses these unmapped reads many times in a way to build confidence in a possible insertion sequence between the reference and the sequenced organism by identifying one leg of the pair mapping to the reference and another that does not. This information in concert with other key coverage characteristics provides these insertion results.

However, if you are ultimately after SNPs and small INDELs the inclusion of this data is not worth it... it only casts doubt on your more confident mappings.

My best.

Jarret Glasscock
Cofactor Genomics
http://www.cofactorgenomics.com