View Single Post
Old 09-22-2011, 07:23 AM   #4
Simon Andrews
Location: Babraham Inst, Cambridge, UK

Join Date: May 2009
Posts: 871

Originally Posted by cascoamarillo View Post
So if I want to do an alignment where there's no limit in the number of hits (-m has no limit) a read is reported (i.e., take each contig in the ref file as independent), should I take off the --best option? Or what do you recommend?
If you want to report multiple alignments you will need to increase -k (number of hits to report) to a suitably large value, and not set -m. We would normally then recommend running this in combination with --best and --strata which will ensure that all of the hits reported are equally good (ie same number of mismatches), so poorer valid hits are not reported if a better hit exists. This combination of options will still suffer from the effect you mentioned earlier where you won't get the same results from mapping against each contig individually compared to mapping against them all together.

If you really want to see every possible valid hit then you'd need to set -k to a high value, not set -m and not set either --best or --strata. This will produce consistent results (as long as you don't reach the limit for -k), but will also include sub-optimal hits when a better hit is found (eg you might see hits with 3 mismatches reported where there is also an exact match in your search set).
simonandrews is offline   Reply With Quote