<30% of reads mapped

[shi]

I tried using bowtie (0.12.2) to mapping Illumina GAII reads to human genome. However, the percentage of reads which were mapped (reads with at least one reported alignment) is less than 30%. The length of these reads is 100bp.

I also have the GAII read data which included 50bp reads (same sample). More than 80% of reads can be mapped using bowtie.

I used the default parameters to run bowtie. Should I adjust the parameter for aligning 100 bp or bowtie is not suitable for mapping reads longer than 50 bp?

I notice that in Table 3 of the bowtie paper the percentage of mapped reads for 76 bp reads is 44.5% (31.7% if using option -v 2), which is much lower than the percentages for 50bp reads and 36 bp reads in the same table.


Thanks,
Wei

[krobison]

bowtie has options (--trim5 & --trim3) for removing N bases from the 5' or 3' end of read -- have you tried this to truncate your reads down to 50nt & see if you get much mapping.

You might also take 10-20 reads or so with high quality scores & BLASTN them (easy to do at NCBI) against all known sequences -- this would check for a sample swap (to another species).

[dcfargo]

I'd drop the -v go with -n (perhaps make -l a little longer than default) and 'titrate' the -e value using a subset of the sequences.