sensitivity/accuracy of fast then sensitive search


New member
ac-diamond has a "sensitive1" mode, which is just an initial search using the 'fast' method, and then for all queries lacking a hit, query just those using the 'sensitive' mode. Do you have any idea what the sensitivity/accuracy (or resource utilization) of diamond when using this sort of 2-pass approach?

Benjamin Buchfink

Staff member
I haven't tried this approach and don't really have any numbers for it. It should work quite well when only looking for the best hit for each query, but it also depends on the data. If say >80% of your queries have best hits with >60% identity, this should work well and improve the performance quite a lot. You will lose some accuracy compared to just running Diamond-sensitive on everything, because it's possible that Diamond-fast does in fact not find the best hit, but a suboptimal one. I would estimate that this effect is pretty small though.