Mmseqs2¶

General information¶

Installed versions¶

Read information about how to load this software in your environment by searching for Lmod module.

MMseqs2 (Many-against-Many sequence searching) is a GPU-accelerated software suite designed for searching and clustering very large protein and nucleotide sequence datasets. It can run up to 10,000 times faster than BLAST, while still maintaining nearly the same sensitivity at 100-fold speed. Additionally, it performs profile searches with sensitivity comparable to PSI-BLAST, but at more than 400 times the speed. More information at https://github.com/soedinglab/MMseqs2

Types of alignment modes¶

By default MMseq2 just finds candidate matches using k-mers and fast heuristics, computes scores and E-values but does NOT compute a residue by residue alignment. If you want full Smith-Waterman alignment which is comparable as used by blast you should add parameter --alignment-mode 3 although this type of mode is considerably slower.

How to run¶

For running the alignment there are a couple of pre/post commands that need to be executed.

Creating your target¶

First of all you need to create a FASTA format file with the sequence you want to search for. This file needs to be converted to adapt to MMseq2

mmseqs createdb [fasta source] [seqDB]

Creating human readible results¶

You need to convert the results in human readibly format which is similar to the blast output

mmseqs convertalis [seqBD] /sw/data/MMseqs2_gpu_data/NR [results_nr] [hits_with_sequences.tsv] --format-output "query,target,evalue,bits,pident,alnlen,qaln,taln"

This will create one file [hits_with_sequences.tsv] containing the results, with alignment.