Low mapping of reads to trinity assembly

[horvathdp]

Hi all,

I am using iPlant resources, and I just finished doing an assembly of my RNAseq data using trinity (great results- ~200K contigs with N50 of ~1500, and mean of ~900- CEGMA giving 241 of the 248 conserved sequences as complete). With this in hand I used tophat2 to map my fragments as the first step to ID deferentially expressed transcripts. However, I am only getting about 38% of my fragments mapping to the Trinity contigs for any given sample. Is this expected? Is there a better way to do this? Any advice would be appreciated.

Dave

[flyingoyster]

I had similar problem, but even lower mapping rate (about 2%). What could be the problem? If you figure it out, please let know. Thanks!

[horvathdp]

More thoughts

I think one of my problems is that at least 25% of the reads are too short to allow mapping of the paired ends. I am running TopHat2-SE to see if that significantly improves the percentage of mapped reads. I also assume that I lost any reads that didn't form contigs of at least 200 bases. I am not sure how to figure out what percentage that is. Any ideas from the experts out there?

Finally, one more question for the experts: I plan on using the tuxedo suite to analyze my data- Mapping my trimmed sequences to the trintity contigs with tophat, usng Cuffmerge and Cuffdiff to ID differentially expressed genes. Is this a reasonable approach? Are there better approaches for mapping back to a denovo assembled transcriptome? Are there any problems I will likely encounter? Any advice will be appreciated.

[ddb]

What is the length of your reads and expected insert size / fragment length? The tuxedo suite is used when you have a genome sequence as reference. I do not think it is the right tool for denovo assembled contigs. For analyzing contigs I would look at Bowtie/Bowtie2 for mapping - RSEM/eXpress for abundance estimation and then one of the bioconductor tools for differential expression testing. I am sure others can suggest different tools. You could try following the Trinity protocol for downstream analysis.