I was hoping this question had been asked, but I haven't been able to find it.  I want to output the unmapped reads from bowtie as a fastq file for subsequent mapping to other genomes (i.e. the "--un <filename>" option).  I know I can extract the unmapped reads by filtering on the bitwise values in the sam output and converting to fastq with the Picard tool, but I'm using colorspace data and bowtie converts them to letterspace. My understanding (coming mostly from forums and personal discussions) was that the color-to-letter conversion was somehow lossy so mapping the colorspace data directly is always preferable.