[Genome] BLAT & microreads

[Ewan Birney]

On 11 Aug 2007, at 07:14, Galt Barber wrote:

>
> Hi, Ming.
>
> See my mail in the pipermail archives to Chris on July 31st about
> blat and minScore. It appears that minScore can't really be raised  
> above
> about half of the qSize.  So if you specify a minScore
> higher than qSize/2, you will still get values as low as qSize/2.
>
> I recommend using pslCDnaFilter or pslReps to post-filter
> psls, these utilities have many useful options.
>
> Comparing to BLAST is probably worthwhile.
>

The other tool which is worth checking out in terms of parameterisation
is exonerate:

    http://www.ebi.ac.uk/~guy/exonerate/

This has both similar algorithms as BLAST, but also things like  
Genewise,
but much much faster.

The really nice thing about exonerate is that you can change nearly any
aspect of its initial matching scenario - the word size, the number of
mismatches, the threshold for considering a new hit. In addition, in
plays extremely well with farm set ups, allowing one on the commandline
to specify a particular subset of sequences in a file to search,
(these are the --querychunktotal and --querychunkid options). Finally
its output is also very flexible, allowing one to write one's own
output format in a "sprintf" like manner.


Guy Slater (guy at ebi.ac.uk) wrote this, and it is really worth using in
high compute sequence matching scenarios.



> -Galt
>
>
> On Fri, 10 Aug 2007, Minghsun Liu wrote:
>
>> Hi,
>>
>> I am wondering if anyone has been using BLAT for mapping  
>> sequencing reads
>> (microreads) produced by Illumina/Solexa or ABI (currently between
>> 26-32mers, with up to 2 mismatches)?
>>
>> After reading some of the previous posts regarding short reads, I  
>> have tried
>> varying several parameters such as tileSize, stepSize, and  
>> minIdentity. I
>> tried specifying minScore as well but it does not seem to work  
>> (i.e. I am
>> still getting hits that scored below the specified minScore). It  
>> seems to
>> work fairly well for my purpose  (using oneOff=1, minMatch=1,  
>> minIdentiy=92,
>> default tileSize and stepSize) but I don't have any good idea  
>> about how to
>> gauge the sensitivity except comparing the results with the that  
>> from using
>> BLAST. Any thoughts?
>>
>> Cheers,
>>
>> - Ming
>> _______________________________________________
>> Genome maillist  -  Genome at soe.ucsc.edu
>> http://www.soe.ucsc.edu/mailman/listinfo/genome
>>
> _______________________________________________
> Genome maillist  -  Genome at soe.ucsc.edu
> http://www.soe.ucsc.edu/mailman/listinfo/genome