Improve speed by advanced filtering about swarm HOT 1 CLOSED

Filtering using k-mers of length 5 have been implemented in SWARM 1.2.0. This eliminates more than 99.9% of the time-consuming alignments and leads to greatly improved speed. The speedup can be more than 100-fold vs the old version using a single thread with a large set of sequences. Parallelisation on multiple threads only becomes efficient with large datasets. A dataset of 312 503 unique sequences of average length 382 took almost two hours on 16 threads with SWARM version 1.1.0. With version 1.2.0 the same dataset can be processed in less than four minutes with 16 threads and less than 10 minutes with a single thread.

from swarm.