Tophat2 high discordant alignments

[ea11]

Hi,

I am mapping paired end RNAseq data using tophat2, but the alignment summary generated is showing I am getting a very high discordant alignment rate. The only tophat options I am specifying is -p 16 and -o "DIR". Below is the output from tophat2:

Left reads:
          Input     :  88556961
           Mapped   :  76938162 (86.9% of input)
            of these:  20429665 (26.6%) have multiple alignments (622137 have >20)
Right reads:
          Input     :  88556961
           Mapped   :  75252663 (85.0% of input)
            of these:  20114304 (26.7%) have multiple alignments (621700 have >20)
Unpaired reads:
          Input     :     68008
           Mapped   :     56927 (83.7% of input)
            of these:      8045 (14.1%) have multiple alignments (9 have >20)
85.9% overall read mapping rate.

Aligned pairs:  65389463
     of these:  18622479 (28.5%) have multiple alignments
                61775607 (94.5%) are discordant alignments
 4.1% concordant pair alignment rate. 

The flagstat output I get is also below:

341377625 + 0 in total (QC-passed reads + QC-failed reads)
189129873 + 0 secondary
0 + 0 supplimentary
0 + 0 duplicates
341377625 + 0 mapped (100.00%:-nan%)
152190825 + 0 paired in sequencing
76938162 + 0 read1
75252663 + 0 read2
263998 + 0 properly paired (0.17%:-nan%)
130778926 + 0 with itself and mate mapped
21411899 + 0 singletons (14.07%:-nan%)
116104620 + 0 with mate mapped to a different chr
80799382 + 0 with mate mapped to a different chr (mapQ>=5) 

I am using cutadapt to remove adapters and remove low quality reads and that is running fine. But then when I pass the paired files onto tophat, the results don't seem good. From what I have read, it is to do with the mate pairs no longer being in sync in the two fastq files. Is there a way around this and to get the number of discordant alignments down?

I have tried aligning the fastq files with tophat2 without passing the files through cutadapt first and the alignment is fine and there is a very low discordant alignment rate, so I'm guessing the fastq files are good, but something is happening after the cutadapt step.
Just as a note, I am not using Galaxy for the analysis.

Thanks

[GenoMax]

From what I have read, it is to do with the mate pairs no longer being in sync in the two fastq files. Is there a way around this and to get the number of discordant alignments down?
Thanks

Use a paired-end aware trimmer like trimmomatic/BBDuk (from BBMap) which keep the paired end files in sync post trimming.

That said, if you are happy with the cutadapt results and just want to fix the PE read order you can do so by using repair.sh from BBMap (paired end reads in two files example): http://seqanswers.com/forums/showpos...0&postcount=45

[ea11]

Thanks for the reply. I though cutadapt did that with the -p option to specify paired end data.
I shall give BBDuk a try and see the results. I was not happy with the results of trimmomatic on my data, so staying away from that trimmer for now.

Thanks

[GenoMax]

Just checking. You are not switching the R1/R2 files when you use them as input for tophat by mistake? That will produce discordant results for obvious reasons.

[ea11]

Nope I am not. R1 files are before the R2 files in the script.

[GenoMax]

BBMap will do spliced alignments so after you use BBDuk you may want to give BBMap a try on the side while you do your TopHat2 runs.

[ea11]

Thanks, I shall have a read and see what the results look like with BBDuk/BBMap while my tophat jobs are running