Why reads in unmapped.bam still align to reference genome?

[SpreeFu]

Hi all,
I got less than 10% reads in unmapped.bam by tophat (Arabidopsis RNA-seq).
Then I took some of them and did blast on NCBI. I expected to see those reads would align to some other species, however, all the reads I tried were still Arabidopsis mRNA. This confused me! Anyone has clue for it?
THANKS!

[dpryan]

Where they paired-end reads and you blasted just one end of the pair? Did the blast results not always have the reads mapping from end-to-end? There are a lot of possibilities for why this happens, the most common being that the reads weren't adapter trimmed.

[mastal]

Perhaps the reads had too many mismatches in the seed region, or whatever cutoff parameters were set for the Tophat alignment.

[SpreeFu]

Perhaps the reads had too many mismatches in the seed region, or whatever cutoff parameters were set for the Tophat alignment.

Hi mastal,
I agree that some of them just contained too many mismatches, but is it possible that SNP exist in thses reads?
here is the command I used:
tophat -p 16 -G genes.gtf -o SP1_thout genome SP1_R1.fq SP1_R2.fq;

[SpreeFu]

Where they paired-end reads and you blasted just one end of the pair? Did the blast results not always have the reads mapping from end-to-end? There are a lot of possibilities for why this happens, the most common being that the reads weren't adapter trimmed.

Hi dpryan, thanks for your reply.
the reads I used were all trimmed. You mean if only one paired-end read mapped, then all of this pair will be put into unmapped.bam?

[mastal]

Hi mastal,
I agree that some of them just contained too many mismatches, but is it possible that SNP exist in thses reads?
here is the command I used:
tophat -p 16 -G genes.gtf -o SP1_thout genome SP1_R1.fq SP1_R2.fq;

Sure, your sample could have SNPs that are different from the reference genome, and these would count as mismatches.

But, for the unmapped reads that you checked with blast, what were the blast alignment stats like? For example: read length, alignment length, %identity, number of mismatches, number of gaps. Would you expect tophat/bowtie to align something with similar stats?

[SpreeFu]

Sure, your sample could have SNPs that are different from the reference genome, and these would count as mismatches.

But, for the unmapped reads that you checked with blast, what were the blast alignment stats like? For example: read length, alignment length, %identity, number of mismatches, number of gaps. Would you expect tophat/bowtie to align something with similar stats?

OK, I get what you mean, THANKS!

[ysnapus]

I looked into some of my unmapped.bam files, and some of the reads align with on the whole length with 100% identity to the genome. It seems to me to be a bug in Tophat, but I would like to see it confirmed by someone else. I have single ended reads, so it is not because of some issues with the mate read aligning.