Assessing a 454 run?

[dan]

Hi all,

We recently got data from a 1/8th of a 454 run. The read length shows
the typical distribution that I have seen at various meetings (see
attached image). However, how should I go about assessing the 'overall
quality' of the run (if such a clear cut concept exists)...

So far I have plotted the distribution of quality per base and the
distribution of mean quality per read. Of course the qualities will
never be 'perfect', but without any experience or any other reference,
I don't know what kind of distributions I should be looking for. i.e.
we see about 20% of all bases with a quality score below 20... is that
a) as good as we are likely to get, b) not bad, c) woah! ask for 20%
of your money back ;-)

It would be great to get any feedback from the experience on the forum.


Note that we do not have a reference genome to align the reads to, but
we do have a reasonable coverage of the chloroplast DNA, and a
reference for that (estimated 2-4 % chloroplast contamination by read,
giving approximately 10x coverage). What is a good tool to identify
SNPs between our read data and that reference? (If I can first
identify the SNPs, I can then estimate the per base error rate using
the reference).

(Actually I found I can do this with MAQ, but I'll leave the question in in case there are alternative suggestions).


Thanks very much for any information,
Dan.

[jnfass]

you might try Gabor Marth's lab's tools (http://bioinformatics.bc.edu/marthlab/Main_Page) ... use Mosaik to align the reads to the reference, and GigaBayes (evolution of their polyBayes tool) to call SNPs from that alignment. In my recollection, it gives you some better control over whether you're looking for SNPs between homozygous or heterozygous individuals, many individuals, etc, and has sound statistical underpinnings to its algorithms.

~Joe