more counts than total library size ?

[Anna Nad]

Hello,
I have mRNASeq data displaying counts per splice variants of mRNA, and I was just wondering how it is possible that the total number of counts in the sample is much higher than the total number of mapped reads (and of total number of reads generated by Illumina).
thank you in advance for your answer,
anna

[aliceb]

Hi Anna,

It sounds like some of your reads are mapping to more than one place. This way they would be reported more than once in your counts.

How are you doing the mapping?

-Alice

[Anna Nad]

Hello Alice,
The tool BWA is used to map the reads with a maximum set at 2 mismatches in the first 32 bases of the sequences, and a maximum of n mismatches in total.
In my sample, library size was about 277 million reads and mapped reads were about 192 million (69,5%). But total counts is about 861 million reads.
So mismatches were the problem? so 192 million are mapped, but it does not mean uniquely mapped ?
thank you Alice,
anna

[aliceb]

Hi Anna,

I'm afraid I haven't used bwa, but we see similar results with bowtie2 (although not quite to the same extent as you).

When we recover counts from our sam file (we're using ht-seq-count), the number of non-unique reads is reported, which we interpret as reads that are counted more than once because they are mapping to more than one place.

I would try your alignments allowing for fewer mismatches and see what happens.

Good luck.

[timydaley]

BWA does report secondary alignments. This is likely the cause, you can remove all but the primary alignment with samtools view -F 0x0100.

[Anna Nad]

thank you Timydaley