TopHat optimization

[mcarnes]

Hi everyone,

New here. Thanks in advance for any replies to my question.

I used TopHat to align 50-bp paired-reads sequences from Human samples. My top most expressed transcript was a “novel 50 bp fragment.” This was obviously incorrect and after much digging I found that it went back to a common read that was misaligned.

The read was incorrectly mapping 14mb away from its mate, but only had 1 mismatch. The correct location (next to its mate) has 3 mismatches. I clearly do not have the optimal parameters selected.

I originally used:

tophat -G genes.gtf --mate-inner-dist 250 --mate-std-dev 100 --b2-very-sensitive --no-mixed --library-type fr-unstranded genome read1_R1.fastq read2_R2.fastq

I would like to give preference to correct mate-pair orientation OVER number of mismatches.

I am considering changing the following:

1) Changing --b2-very-sensitive to --b2-sensitive
2) Removing --no-mixed and/or adding --no-discordant
3) Increasing --read-mismatches to 3


Has anyone experienced this issue or have suggestions on how to optimize the parameters?

Thanks!

[barkasn]

The fact that you have a common read that gives you a false positive suggests you have a high duplication rate. Have you QC'd your data?

I would would start running with default parameters first. It is very easy to spend alot of time optimizing parameters without quite knowing what the effect really is.