Split fastq files for tophat analysis

[Bobbieshaban]

Hi,

Does anyone see anything wrong with splitting of fastq files for alignment to tophat then to merge them together afterwards?

The reason why I want to split them is to be able to make greater use of the cluster we have available.

I am able to split the fastq files using an algorithm I created in perl, the merging of the files seems to work except I am getting a few missing reads when I compare the merged output from my split fastq as compared to when I run the file in tophat as a whole.

For example the split paired end tophat run produces a samtools flagstat of

$ samtools flagstat merged_accepted_hits.bam
37716745 + 0 in total (QC-passed reads + QC-failed reads)
0 + 0 duplicates
37716745 + 0 mapped (100.00%:nan%)
37716745 + 0 paired in sequencing
19017603 + 0 read1
18699142 + 0 read2
35853292 + 0 properly paired (95.06%:nan%)
35974826 + 0 with itself and mate mapped
1741919 + 0 singletons (4.62%:nan%)
0 + 0 with mate mapped to a different chr
0 + 0 with mate mapped to a different chr (mapQ>=5)

While the full fastq filed paired end run from tophat produces

$ samtools flagstat accepted_hits.bam
37739551 + 0 in total (QC-passed reads + QC-failed reads)
0 + 0 duplicates
37739551 + 0 mapped (100.00%:nan%)
37739551 + 0 paired in sequencing
19028732 + 0 read1
18710819 + 0 read2
35896074 + 0 properly paired (95.12%:nan%)
36017796 + 0 with itself and mate mapped
1721755 + 0 singletons (4.56%:nan%)
0 + 0 with mate mapped to a different chr
0 + 0 with mate mapped to a different chr (mapQ>=5)

The difference is only 0.06% of properly paired reads, but may be missing some useful information. I have checked the splitting of the files and the numbers of the lines are exactly the same.

http://seqanswers.com/forums/showthr...at+fastq+split. this thread suggests that some "low abundance splice sites" are lost.

Would anyone have anymore information about this?

Thanks for the help,

Bobbie.

[dGho]

split files

I have split fastq files to run Tophat. From what I understand is that this is a fairly common practice. Here is a hypothetical example:

#split read 1 into smaller files after every 40,000,000 lines
split -l 40000000 wholefile_read1.fastq ;
#rename resulting files
mv xaa wholefile_read1_1.fastq
mv xab wholefile_read1_2.fastq
.
.
#split read 2 into smaller files after every 40,000,000 lines
split -l 40000000 wholefile_read2.fastq
#rename resulting files
mv xaa wholefile_read2_1.fastq
mv xab wholefile_read2_2.fastq
.
.
#align split files with tophat
tophat -o out_1 -G mm10.gtf mm10 wholefile_read1_1.fastq wholefile_read2_1.fastq
tophat -o out_2 -G mm10.gtf mm10 wholefile_read1_2.fastq wholefile_read2_2.fastq
.
.
#use samtools to put the bam files back together
Samtools merge out.bam out_1 out_2

[dGho]

I didn't answer your question

I guess I did not exactly answer your question though. I do not know if there is any difference in results when the files are split. I do know that my very experienced co-worker does it all the time. That does not necessarily help.