RT Journal Article
SR Electronic
T1 Linclust: clustering protein sequences in linear time
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 104034
DO 10.1101/104034
A1 Martin Steinegger
A1 Johannes Söding
YR 2017
UL http://biorxiv.org/content/early/2017/01/29/104034.abstract
AB Metagenomic datasets contain billions of protein sequences that could greatly enhance large-scale prediction of protein functions and structures. Lin-clust can cluster sequences down to 50% pairwise sequence similarity and its runtime scales linearly with the input set size, not nearly quadratically as in conventional algorithms. We cluster 1.7 billion sequences from ~2200 metagenomic and metatran-scriptomic datasets in 30 hours on 28 cores. Free software and data available at https://mmseqs.com/.