Tophat should be used when mapping reads to the genome, not the transcriptome. Because you're mapping your reads to the transcriptome assembled via Trinity, Bowtie or BWA are good choices. This also changes your downstream analyses, because Cufflinks does not work well on reads mapped to the transcriptome. Tools for quantitating transcriptome-mapped reads include RSEM and eXpress. Good luck, J.
I've been using the Main Galaxy server to work on an RNA-Seq project for a non-model plant, and I've noticed that my output from Tophat and Cufflinks might not be as good as I'd like. I have a reference transcriptome assembled in Trinity, and it is based on the same Illumina-generated 100 bp reads I'm trying to map to it. When I use Tophat to map the reads to the reference transcriptome (I have trimmed the reads and filtered the lower quality ones), only about 10% of the reads actually map, so I go from 30,000,000 reads before mapping to 3,000,000 that are actually mapped. Therefore, I feel like I'm losing a lot of data. When I've changed the parameters to allow for more mismatches, not many more reads seem to map, and in many cases, the Tophat run fails and I receive the error message: "Settings: Output files: "/tmp/3030460.cyberstar.psu.edu/tmpWbxTnm/dataset_5530451.*.ebwt" Line rate: 6 (line is 64 bytes) Lines per side: 1 (side is 64 bytes) Offset rate: 5 (one in 32) FTable chars: 10 Strings: unpacked Max bucket size: def". I've had similar numbers of reads map with Bowtie by itself and BWA as well. I've also tried mapping the reads to the assembled isoforms (contigs) of the transcriptome, and this results in many more reads (close to 90%) being mapped. Therefore, I figure the reads should map to the reference transcriptome, and I'm not sure why this isn't happening.
The other issue I've run into is that in Cuffdiff only about 4,800 genes appear in the output files as being tested for differential expression. There are approximately 100,000 genes in the reference transcriptome, so I was thinking that there should be more than ca. 4,800 that are tested for differential expression. Should each gene be tested? Does Cuffdiff just not report some of the genes that are not differentially expressed, or is the program not testing all of the genes?
If anyone can provide some help, guidance, or a suggestion, I'd greatly appreciate it. Thanks, and take care.
Jim
___________________________________________________________ The Galaxy User list should be used for the discussion of Galaxy analysis and other features on the public server at usegalaxy.org. Please keep all replies on the list by using "reply all" in your mail client. For discussion of local Galaxy instances and the Galaxy source code, please use the Galaxy Development list:
http://lists.bx.psu.edu/listinfo/galaxy-dev
To manage your subscriptions to this and other Galaxy lists, please use the interface at: