Tophat2 Alignment Rate

[SHeaph]

Hi,

I am new to Tophat and I have an alignment rate query.

I am trying to map RNASeq paired reads (100bp) to a reference genome using Tophat (v2.0.6) and I am returning an alignment rate of 26%. The –solexa-quals option improved the alignment rate slightly, altering other options from the default settings had no affect. I have 34,000 reads and I was expecting a higher rate.

tophat -p 2 --solexa-quals --b2-very-sensitive Homo_sapiens.GRCh37.67.cDNA_ncRNA UC07i.qfiltered.Bhg19.1.fq UC07i.qfiltered.Bhg19.2.fq

I used the same reads in Bowtie version 2.0.2 and returned an alignment rate of 90.5%

bowtie2 -p2 -x Homo_sapiens.GRCh37.67.cDNA_ncRNA -1 UC07i.qfiltered.Bhg19.1.fq -2 UC07i.qfiltered.Bhg19.2.fq -S UC07i.sam --very-sensitive-local

I understand that Bowtie2 is not appropriate for mapping RNASeq reads as it does not account for RNA splicing and Tophat uses end-to-end alignment.

Is it possible to improve the alignment rate for Tophat or can I assume that 26% is an appropriate rate?

The adapter sequences have been removed so I don't think trimming would be useful.

Thanks in advance!

[dpryan]

It looks like you're mapping against some subset of the transcriptome. Try mapping against the actual genome with both (I would recommend using a reference GTF file with tophat) and I expect tophat will perform more favourably.

[DunderChief]

I would be cautious about switching to solexa scores just because you get a higher alignment rate. It seems like you have a problem unrelated to your quality scores and you'll probably end up confusing things further. If you run fastQC, it will automatically determine the version of your quality scores.

I doubt this would make that big of a difference anyway, but are you sure you're using tophat v2. It depends on how you set it up on your system, but typically the command is tophat2, not tophat.

Also, how are you determining the alignment rate? When I first started using tophat, I got very confused by their log files. Calculate the % mapped the same way for both tophat2 and bowtie2 results.

[SHeaph]

Thanks for yer help

[SHeaph]

I was using a cDNA and non-coding RNA library to map the reads against. Can I use this as a GTF? if so would any unmapped reads here then be mapped against the actual genome?

Much appreciated.

[dpryan]

You can't use the multifasta file as an annotation (since it doesn't actually annotate anything), but since it's name suggests that it's from Ensembl, you might just use the normal Ensembl genome sequence and annotation.

You can probably save some time by just downloading the premade indices (and I think GTF files, though I don't recall exactly) from here.

BTW, don't be surprised if mapping things this way leads to slightly lower alignment rates, as the results are going to be both more reliable and easier to analyse downstream (at least for common analyses).