mapping 35bp solid reads with tophat unsuccessful

[Ilarius]

Hi, I recently downloaded some solid short reads (35 bp), single end, and convert them in cfasta with ahi-dump.

I found that there are very few mappers that support solid read mapping in color space and finally i decided to use tophat using bowtie1 (which supported color space mapping).
Unfortunately I less less than 30 % of the reads mapping.
My previous experience has always been with 100 bp paired end reads, so I am not completely sure if I am using the right parameters for short read mapping:

tophat -G $annotation --segment-length 17 --segment-mismatches 1 -g 1 --bowtie1 -p 12 -o $OUTPUT_FOLDER --coverage-search --color --quals $genome $READ $QUAL

maybe is the segment-mismatch and maximum multi-hits too restrictive? (-g is 20 by default). What for your experience are the best parameters for short read mapping?

Thanks
Ilario

[colindaven]

In general, in comparison to Illumina PE reads don't expect high mapping percentages. From memory I think 50-70% was pretty good, as the data were not extensively filtered (on the 5500xl machine at least).

You probably don't want to use Tophat for such short reads as I don't think it will be able to find anchors/seeds successfully.

My best experiences with SOLiD were with the commercial program NovoalignCS. You might be able to get a free trial.

Otherwise, Shrimp2 was also not too bad.

Why not use just bowtie1 instead of tophat2 with bowtie1 ?

[Ilarius]

I did use tophat2 with bowtie1, anyway if I let the default -g option (i think it is 20) it takes ages to finish.

As regards shrimp I didn't find any working link to download it!

[colindaven]

I recently found another non-commercial aligner for SOLiD data - it is called subread. I have only tried for Illumina data but it's very good there.

Again, I wouldn't use tophat for such short read data.

[gringer]

What for your experience are the best parameters for short read mapping?

Anything that gets you quicker to re-analysing your target sample with a non-SOLiD system. You shouldn't be mapping SOLiD reads at all, especially not 35bp ones; interpreting the results is just too confusing (even for the people writing code to do the mapping).