Releases: DSIMB/SWORD2
SWORD2 v2.0.1
This release fixes an issue where the new optimized code of Peeling produces different Protein Unit (PU) delineations compared to the old code. The discrepancy was traced back to differences in how the DSSP file is parsed, specifically regarding residue numbering and handling of certain residues.
The webserver is now also updated to this release.
Full Changelog: v2.0.0...v2.0.1
SWORD2 v2.0.0
Major improvements:
Peeling
- Added source code
- Compiled at installation time and is compatible with amd64 and arm64 (Apple Silicon).
- Vastly improved the code to be compliant with C standards and best practices
- Time complexity reduction, refactoring and less redundancy
- Parallelization using OpenMP on critical functions to benefit from multiple cores
- New option to choose the number of CPUs
- Better memory and i/o management
Overall, Peeling is now ~10x faster than before
Scoring
- Parallelization using OpenMP to benefit from multiple cores.
- Compiled at installation time and is compatible with amd64 and arm64 (Apple Silicon).
This program performs the calculation of the pseudo-energies for domains and PUs in order to calculate the Autonomous Unit Likelihood.
SWORD2
- Several minor fixes
- Parallelization of contact probability matrices plots generation to improve overall speed
- Added the option
--disable-energies
to disable calculation of pseudo-energy - Added the option
--disable-plots
to disable generation of contact probability matrices plots - Detect and parse automatically the first chain of the input PDB/mmCIF file if not provided in CLI
SWORD2 is now between 2 and 700x faster than previous version, depending on the protein length.
Full Changelog: v1.1.2...v2.0.0
SWORD2 v1.1.2
Standalone
- New Dockerfile (non-root generated outputs)
- New JSON formatted output
Webserver
- New JSON formatted output
v1.1.1
Features:
- Added a new optional argument
model
. Defines which model of the structure file user wishes to parse. For example in NMR PDB/mmcif structure files, comprising usually several models, user can choose a specific one. By default SWORD2 parses the model 1.
v1.1.0
Bug fixes:
- SWORD2 is now compatible up to macOS 13.0 Ventura Intel, arm64 (M1, M2).
- Added an
install.sh
script to compile a few dependencies for best native performances. - Fixed few typos and wrong paths.
Feature:
- Enable users to query a MGnify protein structure predicted by ESMFold, fetched from the ESM Metagenomic Atlas database released by MetaAI.
Breaking changes:
- Long version command line arguments changed from "*_code" to "*_id" for consistency.
v1.0.0
First release of the standalone version of SWORD2 method.
This release enables the protein 3D structure domain delineation of:
- PDB or mmCIF input file.
- PDB id: the structure file is automatically downloaded from the RCSB PDB database.
- AlphaFold DB Uniprot Accession Id: the structure file is automatically downloaded from the new release of the AlphaFold Protein Structure Database (https://www.deepmind.com/blog/alphafold-reveals-the-structure-of-the-protein-universe).
This release was tested on Linux, macOS (intel, arm64) and Windows 10.
Corresponding docker image is available: https://hub.docker.com/r/dsimb/sword2