[acadMT]

I have questions regarding to reads from Solid 5500 mapping.

Dear all,
I have recently been working on dataset (75bps single end reads/fragment) output from solid 5500. I used bowtie to map to mouse genome, but the result was very unreasonable. There were only about 10 percent of total reads could be mapped to mouse genome.

Then, I used blat from UCSC to blat the unmappable reads back to mouse genome. I came across that unmappable reads from bowtie output had many output after I "blatted". Then, I took a look at the result from blat, I noticed that most of them could map/align to the mouse genome with shorter length. For instance, the original 75bps read that could not map to mouse genome with bowtie, I got some output with shorter reads (46 bps) from blat. I wonder where the rest of the reads went?

p.s: almost all the unmappable reads from bowtie got output from blat with shorter reads length aligned to mouse genome.

Mike

[pmiguel]

Are you mapping in color space? If not, were these ECC reads, or just normal chemistry? If at all possible, you want to do your mapping in color space. Bowtie allows you to do this.

Have you run the data set through a QC checking program? Eg, FastQC? Might be worth it to see if the quality dropped off dramatically at some point in the run.

--
Phillip

[acadMT]

The csfasta and qual file obtained from the solid XSQ barcode separation process, then color space was converted to nucleotide (convert the file to fastsanger format in galaxy then used "manipulate FASTQ" tool in galaxy to convert to NT.)

They were not ECC reads, it was just normal chemistry which followed ABI protocol. By the way, is it normal to amplify to 2^20 cycle for a library ? thanks in advance.


Mike

[Chipper]

For instance, the original 75bps read that could not map to mouse genome with bowtie, I got some output with shorter reads (46 bps) from blat. I wonder where the rest of the reads went?

They went off into colorspace. If you absolutely need to map your reads in basespace you could try bowtie2 with the --local option, but mapping in colorspace is the only way to get reasonable results.

[colindaven]

Hi Mike,

as others have said map in colour space using the csfasta and qual files. Alternatively to bowtie you could try bwa, though both don't seem great for 60 bp data from the 5500xl (33% and 40% mapping rate for human SE data). It seems the bwa developers are considering stopping their

Bioscope and Clc Genomics workbench seem a lot better (~85 % mapping rate), in part because they (or at least Bioscope) iteratively trim the sequences.

Novoalign-CS is also supposed to be very promising, but we are just in the process of getting a license.