Considering Quality scores of reads when aligning

[mattanswers]

In the following URL the is a graph relating Q score (Sanger and Solexa) with p-value: http://en.wikipedia.org/wiki/FASTQ_format
Basically, I think anything above a Q (solexa) score of 20 is very acceptable. From 20-13 the probability begins to vary much more. Around a Q score of 13 it seems that there is a 0.05 chance of a bad call.

[maubp]

I'm not sure what exactly TopHat would do in this situation.

You could try converting your reads from the Illumina 1.3+ FASTQ file format (aka phred64) to a Sanger FASTQ file (aka phred33).

There are lots of tools to do this conversion (search the forum), I'm biased but would suggest EMBOSS seqret for a command line tool, or for a script based solution BioPython (use function Bio.SeqIO.convert for this) or BioPerl etc.

[thinkRNA]

I have reads with quality scale on phred64. I get an error when using --solexa1.3-quals option in tophat which is a known error (Error: could not execute prep_reads)

So, since I can't use this option, does it mean my reads get aligned without the quality scores being taken into consideration by bowtie?

Finally, how does one decide what is a good quality score? What if there are really bad quality reads in the seed region (beginning of the read) but good ones towards the end giving it a high quality score. In this case, I would like to throw this read .

Any one has any thoughts on threshold used?