SAbR (Structure-based Antibody Renumbering) renumbers antibody PDB files using the 3D coordinate of backbone atoms. It uses custom forked versions of SoftAlign and ANARCI to align structures to SAbDaB-derived consensus embeddings and renumber to various antibody schemes, respectively.
Full API documentation is available at sabr.readthedocs.io.
Requirements: Python 3.11 or higher
- SAbR can be installed into a virtual environment via pip:
# Latest release
pip install sabr-kit
# Most recent version from Github
git clone --recursive https://github.com/delalamo/SAbR.git
cd SAbR/
pip install -e .It can then be run using the sabr command (see below).
- Alternatively, SAbR can be directly run with the latest docker container:
docker run --rm ghcr.io/delalamo/sabr:latest -i input.pdb -o output.pdb -c CHAIN_IDPractical considerations:
- Heavy and light chain structures are similar enough that chain type should be manually declared with
--chain-typeif possible (leave blank if uncertain). - It is recommended for now to truncate the query structure to contain only the Fv when running SAbR, as it will sometimes align variable region beta-strands to those in the constant region.
- When running scFvs, it is recommended to run each variable domain independently.
If running on a Mac with apple silicon, set the environmental variable JAX_PLATFORMS to cpu.
Usage: sabr [OPTIONS]
Structure-based Antibody Renumbering (SAbR) renumbers antibody structure
files using the 3D coordinates of backbone atoms. Supports both PDB and
mmCIF input formats.
Options:
-i, --input-pdb FILE Input structure file (PDB or mmCIF format).
[required]
-c, --input-chain TEXT Chain identifier to renumber (single
character). [required]
-o, --output FILE Destination structure file. Use .pdb
extension for PDB format or .cif extension
for mmCIF format. mmCIF is required for
antibodies with extended insertion codes
(e.g., very long CDR loops). [required]
-n, --numbering-scheme [imgt|chothia|kabat|martin|aho|wolfguy]
Numbering scheme. [default: IMGT]
-t, --chain-type [H|K|L|heavy|kappa|lambda|auto]
Chain type for ANARCI numbering.
H/heavy=heavy chain, K/kappa=kappa light,
L/lambda=lambda light. Use 'auto' (default)
to detect from DE loop occupancy.
[default: auto]
--overwrite Overwrite the output file if it already
exists.
-v, --verbose Enable verbose logging.
--residue-range START END Range of residues to process in PDB
numbering (inclusive). Use '0 0' (default)
to process all residues. Example:
--residue-range 1 120 processes residues
1-120.
--random-seed INTEGER Random seed for JAX operations. If not
specified, a random seed will be generated.
Set this for reproducible results.
--disable-deterministic-renumbering
Disable deterministic renumbering corrections
for loop regions. By default, corrections are
applied for FR1, DE loop, and CDR loops.
--disable-custom-gap-penalties Disable custom gap penalties for alignment.
By default, custom penalties are applied
including: zero gap open penalty in CDR
regions (IMGT 27-38, 56-65, 105-117), zero
gap open at position 10, and overhang
penalties at sequence termini.
-h, --help Show this message and exit.SAbR can also be used programmatically to renumber BioPython Structure objects directly in memory:
from Bio.PDB import PDBParser, PDBIO
from sabr import renumber
# Load a structure
parser = PDBParser(QUIET=True)
structure = parser.get_structure("antibody", "input.pdb")
# Renumber the structure (returns a new BioPython Structure)
renumbered = renumber.renumber_structure(
structure,
chain="H", # Chain identifier
numbering_scheme="imgt", # imgt, chothia, kabat, martin, aho, wolfguy
chain_type="auto", # H, K, L, or auto
)
# Optionally specify a residue range
renumbered = renumber.renumber_structure(
structure,
chain="H",
res_start=1, # Start at residue 1
res_end=128, # End at residue 128
)
# Save the renumbered structure
io = PDBIO()
io.set_structure(renumbered)
io.save("output.pdb")