unique mapping for chip-seq?

[frozenlyse]

Ok so I am re-analysing some publicly available chip-seq data (human H3K4me3 & H3K27me3) and have a few questions

at the moment my "pipeline" is to align reads with bowtie (using --best), convert to maq format and generate .wigs & peaks with findpeaks

Comparing to the publication, I am mapping 29M/40M reads, where they only mapped ~9M so i believe they are throwing out reads which map equally well to multiple places in the genome, whereas bowtie is reporting a random sampling of the best alignments (ala maq).

does anyone have thoughts on which way is better to go for chip-seq peak finding?

[Chipper]

What setting did you use for bowtie (-n or -v) and what read lengths was it. I think default for bowti is to use up to 2 mm in the good part of reads (first 28 bases or so), -v 2 would mean up to 2 mm in the full sequence.

Probably setting a relatively low -m, or only use reads with multiple matches if they are of high quality would be best since it may be biased to report random repeat matches otherwise. But it would be very interresting to see how it affects the k27 signal since it could be present over lare distances including repeaats.

[frozenlyse]

my read lengths are 36bp, and i used the default (i think this is -n 2 but im not sure!)

i am rerunning my analysis with -m 1 to see how that changes the analysis, ill post back with the results!