Initial QC and grooming for Illumina HiSeq2000 paired end RNAseq on Galaxy

[lindseykelly]

I am trying to do RNAseq analysis on Paired end data from the Hiseq2000. I have about 50 files for each sample (25 forward and 25 reverse - although each sample has a different number of files).

I think that I need to:
-convert them into FASTQ sanger format using the FASTSQ groomer tool
-check the quality using the FASTQqc tool

I don't know how to handle this many files. Do I have to groom and run the QC for each file? Should I join the paired files and run both tools on each pair, or should I combine all of the data for each sample (which I don't know how to do) and then groom and run the QC for all of the reads for the sample.

Thanks in advance for advice
Lindsey

[lindseykelly]

This was the response from the Galaxy team, in case someone else has this question:

Yes, you have this correct. The general path would be to:

- join forward and reverse data per run
- run FASTQ Groomer & FastQC
(note: if your data is already in Sanger FASTQ format with Phred+33 quality scaled
values, the datatype '.fastqsanger' can be directly assigned and the FASTQ Groomer
step skipped. This is likely true if your data is a from the latest CASAVA pipeline, but
please double check.)
- discard data as needed based on quality
- split forward and reverse data that passes QC
- concatenate all forward reads from a sample into one FASTQ file
- concatenate all reverse reads from a sample into one FASTQ file.
- for each sample, run TopHat using the two concatenated FASTQ files

To manipulate paired end data, please see the tools -> NGS: QC and manipulation: FASTQ splitter & FASTQ joiner.

To combined data files head-to-tail from multiple runs into a single FASTQ file please see the tool -> Text Manipulation: Concatenate datasets.

I am not sure of the actual volume of data, but if these start to get large or TopHat errors with a memory problem, a local or cluster instance would be the recommendation: http://getgalaxy.org

For reference:

http://tophat.cbcb.umd.edu/manual.html

http://www.nature.com/nprot/journal/v7/n3/full/nprot.2012.016.html

Hopefully this helps. Others are welcome to post comments/suggestions.

Jen
Galaxy team

[[email protected]]

This may be helpful to you:

http://www.illumina.com/documents/products/datasheets/datasheet_rnaseq_analysis.pdf

[mhkiani]

Broken paired reads

I got some RNA-seq paired 100bp data and when I did the RNA-seq analyis with CLC, I got more than 50% broken pairs among the reads and I'm not sure why.

[sugo]

What is the purpose of joining the forward and reverse reads prior to QC? Couldn't the QC be run on the separate reads?

[Mike2188]

If you do each file individually then you run into errors during alignments. For instance if I had 100,000 paired end reads in two files forward.fq and reverse.fq and I performed some trimming and quality filtering on each individually, I might end up with one file with 90,000 reads and one with 89,000. Now when I go to do alignments, the program will assume the first read in forward.fastq corresponds to the first read in reverse.fastq - but now the files are uneven. The alignments won't work correctly because of this.