Foldmason

Foldmason builds multiple alignments of large structure sets.

Publications

van Kempen M, Kim S, Tumescheit C, Mirdita M, Lee J, Gilchrist C, Söding J, and Steinegger M. Fast and accurate protein structure search with Foldseek. Nature Biotechnology, doi:10.1038/s41587-023-01773-0 (2023)

Table of Contents

• Foldmason
• Webserver
• Installation
• Documentation
• Quick Start
  • Multiple alignment
    • Output
    • Important Parameters
  • Databases
    • Create Custom Databases and Indexes
• Main Modules
• Examples

Webserver

Align your protein structures quickly using our Foldmason webserver: ...

Installation

# Linux AVX2 build (check using: cat /proc/cpuinfo | grep avx2)
wget https://mmseqs.com/foldmason/foldmason-linux-avx2.tar.gz; tar xvzf foldmason-linux-avx2.tar.gz; export PATH=$(pwd)/foldmason/bin/:$PATH

# Linux SSE2 build (check using: cat /proc/cpuinfo | grep sse2)
wget https://mmseqs.com/foldmason/foldmason-linux-sse2.tar.gz; tar xvzf foldmason-linux-sse2.tar.gz; export PATH=$(pwd)/foldmason/bin/:$PATH

# Linux ARM64 build
wget https://mmseqs.com/foldmason/foldmason-linux-arm64.tar.gz; tar xvzf foldmason-linux-arm64.tar.gz; export PATH=$(pwd)/foldmason/bin/:$PATH

# MacOS
wget https://mmseqs.com/foldmason/foldmason-osx-universal.tar.gz; tar xvzf foldmason-osx-universal.tar.gz; export PATH=$(pwd)/foldmason/bin/:$PATH

# Conda installer (Linux and macOS)
conda install -c conda-forge -c bioconda foldmason

Other precompiled binaries for ARM64 amd SSE2 are available at https://mmseqs.com/foldmason.

Quick start

Multiple alignment

The easy-msa module allows you to align multiple query structures formatted in PDB/mmCIF format (flat or gzipped). By default it outputs the alignment as a FASTA-format file as well as an interactive HTML output.

foldmason easy-msa <PDB/mmCIF files> result.fasta tmpFolder

Output

FASTA alignment

Foldmason generates alignments in FASTA-format, with both amino acid and 3Di alphabets (_aa.fa and _3di.fa suffixes, respectively).

Interactive HTML

Foldmason can generate a HTML MSA visualisation using the msa2lddtreport module.

foldmason msa2lddtreport myDb result.fa result.html

This is generated automatically when using the easy-msa. The following will produce result.fasta and result.html.

foldmason easy-msa <PDB/mmCIF files> result tmpFolder

Important parameters

Option	Category	Description
--gap-open	Alignment	Gap opening penalty (default: 10)
--gap-extend	Alignment	Gap extension penalty (default: 1)
--refine-iters	Alignment	Number of refinement iterations to run after initial alignment (default: 0)
--output-mode	Alignment	0: Amino acids, 1: 3Di alphabet (default: 0)
--pair-threshold	Scoring	Maximum proportion of gaps in column threshold for LDDT calculation (default: 0.0)

Create custom databases and indexes

The structure database can be pre-processed by createdb. This make sense if searched multiple times.

foldmason createdb example/ structureDB

Main Modules

• easy-msa multiple alignment workflow from structure files
• structuremsa multiple alignment from structure database
• msa2lddt calculate structure-based score (LDDT) of a MSA
• refinemsa iterative MSA refinement

Examples

Basic MSA workflow

The easiest way to use Foldmason is to use the easy-msa workflow like so:

foldmason easy-msa <PDB/mmCIF files> result tmpFolder

By default, easy-msa produces multiple alignments in FASTA format (result_aa.fa and result_3di.fa for amino acid and 3Di alphabets, respectively), as well as a Newick format tree (result.nw). This is equivalent to the following sequence of commands:

foldmason createdb <PDB/mmCIF files> myDb
foldmason structuremsa myDb result

easy-msa can also compute the average LDDT score of the alignment and generate the interactive HTML visualisation by specifying --report-mode 1, like so:

foldmason easy-msa <PDB/mmCIF files> result tmpFolder --report-mode 1

This is the equivalent of the following sequence of commands:

foldmason createdb <PDB/mmCIF files> myDb
foldmason structuremsa myDb result
foldmason msa2lddtreport myDb result_aa.fa result.html --guide-tree result.nw

Note: the generated guide tree is passed to msa2lddtreport to display it inside the HTML report.

Aligning large data sets

Foldmason can use the clustering capabilities of Foldseek to pre-cluster input structures before alignment by specifying --precluster, allowing for alignments of large sets of proteins.

foldmason easy-msa <PDB/mmCIF files> result tmpFolder --precluster

Computing LDDT of an externally created MSA

The msa2lddt module computes an average Local Distance Difference Test (LDDT) score over the length of an MSA. This can be done automatically in the easy-msa workflow by specifying --report-mode 1, but msa2lddt can be called separately to compute the LDDT of any given alignment, so long as the structures in the MSA are present in the given structure database:

foldmason msa2lddt myDb result.fa
foldmason msa2lddtreport myDb result.fa result.html

Average MSA LDDT is calculated by computing per-column LDDT scores of every pair of sequences in the MSA, and then averaging them over the length of the MSA. Sequence comparison order is determined by database keys, thus scores for MSAs produced by different tools are comparable if specifying the same structure database in msa2lddt.

MSA Refinement

The refinemsa module refines an MSA by iteratively splitting and re-aligning it, saving a resulting MSA when an increase in average LDDT is detected.

foldmason refinemsa myDb result.fasta refined.fasta --refine-iters 1000

Refinement can be run automatically in the easy-msa workflow by specifying the --refine-iters argument.