How mismatches effects alignments in tophat

[nareshmvr]

I mapped my RNA-Seq paired end reads with tophat, on flagstat my results shows very less number of reads mapped to total number of reads submitted , i guess when more mismatches allowed we are supposed to get more number of reads , but here my reads were inconsistent when more mismatches allowed .

Am i right that when more mismatches allowed , we are supposed to get more number of alignments , please help me in understanding how this mismatches effects alignments and what would be the possible reasons that am getting less number of reads , here am sending my results in the attachment.

Thanks in advance

[Brian Bushnell]

I can't explain those results, but since you are getting such a low mapping rate, you may have pretty low quality (or contaminated) data. It's worth BLASTing some of the unmapped reads to see if they are from another species.

If this is a quality problem, I suggest adapter-trimming and quality-trimming (to ~Q10) the reads prior to mapping, and mapping with BBMap, which is much more tolerant of low-quality data.

[GenoMax]

@nareshmvr: I won't call that "very less". It appears that at least 70-79% of your reads are mapping with proper pairing and another (8-13%) are mapping as singletons.

In any case, if you have not trimmed/cleaned the data before alignment then do follow Brian's suggestion.

[Brian Bushnell]

I read that as 12.7 million reads with only ~4 million mapping... did I read something wrong?

[GenoMax]

Nope. I did not read the title, Yikes.

I thought there were four separate sample stats in there. It looks like all four refer to total 12.7 million "raw" starting reads in the title.

Hopefully that is not some kind of contamination.

[nareshmvr]

Thank you very much responding max and brain

I have done preprocessing step with prin-seq and checked the quality with fastqc before and after preprocessing , they gave positive results , here am sending in the attachment .

Max , the attachment results are for the same sample with different mismatches , as you mentioned it may not be because of contamination , please suggest me the possible reasons for less number of reads and inconsistency in reads mapped when more mismatches allowed .

please reply me in this.

[GenoMax]

@naresh: You are going to need to examine some of the unmapped reads by BLAST (or some other means) to see if they belong to your sample or are "contaminants". You can run bowtie2 with --un* option (http://bowtie-bio.sourceforge.net/bo...ie2-options-un) or BBMap with outu= (or outu1=R1.fq outu2=R2.fq, for paired end reads) option to collect the reads that are not mapping to your reference.

[nareshmvr]

@max and brain

i think i can use umappedreads.bam from tophat , as i am using Tophat for mapping, is there anything specific in using bowtie and BBMap , please reply me in this in that case i ll give with either of them.

[Brian Bushnell]

There's no particular reason to use BBMap over Tophat, or look for contamination, if you are satisfied with your low mapping rate and low sensitivity. A lower sensitivity will incur greater bias due to platform-specific errors, and contamination can completely ruin your experiment, but if you consider a 30% mapping rate to be acceptable then go for it. Peer-reviewers will not consider it acceptable, though, so it might be worthwhile to consider why the mapping rate is so low if you intend to publish this data.

[nareshmvr]

Thank you brain , yes i need to know this for my down stream analysis ,BLASTing of unmapped reads worked well , its contamination , thanq very much