TopHat mapping

[Karen Chait Berman]

Hi,
I am mapping tomato samples with TopHat. In order to determine how many mismatches should be allowed per read, three mapping attempts were done for each sample: allowing 1, 3 and 5 mismatches per read, and 1, 3 and 3 mismatches per a segment of a read, respectively.
The command for 3 mismatches for example was:
'tophat-2.0.9.Linux_x86_64/tophat -o 248 -p 1 -N 3 --b2-sensitive --no-coverage-search --read-edit-dist 3 --segment-mismatches 3 -G Tomato_2.30/tomato_2.30.gff3 Tomato_2.30/S_lycopersicum_chromosomes.2.30 Sample_248_trimmed.fastq'

Following are the results for 2 samples (percentage of uniquely mapped reads) of the 3 mapping trials.
1 mis 3 mis 5 mis
269 91.83% 94.12% 93.59%
274 91.74% 94.06% 93.53%

I was very surprised to see a decrease in the percentage of uniquely mapped reads between 3 to 5 mismatches (and increase in the unmapped reads).
Does anyone have an explanation for it?
Thanks,
Karen

[Michael.Ante]

Hi Karen,

Tophat allows each read only to have a limited numbers of alignments. Per default, this number is 20. Thus, if a reads maps more than 20 times, it is discarded.

You increased with the higher allowed number of mismatches also the number of alignments a read could have. This lowers the amount of uniquely mapping reads and total mapping reads.

Cheers,
Michael