Do I need to use tophat2 for human RNA-seq analysis?

[ymc]

I am not interested in finding novel transcripts. Shouldn't it be easily to simply align my RNA reads to known transcripts using a naive mapper (e.g. bwa) as oppose to aligning to the genome using tophat2?

Am I on the right track?

[chadn737]

Tophat does not reconstruct novel transcripts, it finds novel junctions. I would recommend sticking to a splice aware aligner because ideally it will align more reads. Furthermore, any reads that map to junctions not in your transcriptome file may be incorrectly mapped by mapping to the transcriptome. You can speed up Tophat by providing a gff file and building a transcriptome index and having it align to known transcripts first.

After you have aligned the data, if you are only interested in gene level differences, then there are plenty of pipelines that avoid using cufflinks and having to discover new transcripts.

[ymc]

Thanks for your reply and suggestion. I thought the transcriptome index can speed up my run but it didn't. The run time increased from 9.5 hrs to 11.5 hrs. What went wrong? Here are the command I used to create index and the command that uses the created index.

/tank/rnaseq/tophat-2.0.9.Linux_x86_64/tophat2 -p 6 -r 50 -G genes.gtf --transcriptome-index=hg19rna -o P4N /tank/bowtie2/hg19 SRR493945_1.fastq.gz,SRR493945_1.fastq.gz SRR493945_2.fastq.gz,SRR493946_2.fastq.gz

/tank/rnaseq/tophat-2.0.9.Linux_x86_64/tophat2 -p 6 -r 50 --transcriptome-index=hg19rna/genes -o P4NN /tank/bowtie2/hg19 SRR493945_1.fastq.gz,SRR493945_1.fastq.gz SRR493945_2.fastq.gz,SRR493946_2.fastq.gz

[kopi-o]

The purpose of using the transcriptome index is to increase sensitivity, not necessarily to speed up the run (even though it could do that too, I suppose).

If you want to map RNA-seq reads to the human genome faster, you can use STAR instead of TopHat.