AB RMSD

Table of Contents

• About
• Getting Started
• Usage

About

check here for a chinese tutorial.

Calculate the RMSD between two antibody structure (including nanobody and antibody).

Getting Started

First, install abnumber using conda.

conda install -c bioconda abnumber

Then, install ab_rmsd from github.

pip install git+https://github.com/pengzhangzhi/ab_rmsd.git

[Optional] DockQ

If you want to calculate the DockQ between complex structures, run the following command.

cd DockQ
make

Usage

Calculate the RMSD between predicted and native nanobody structure.

from ab_rmsd import calc_ab_rmsd

# two example pdb files are provided in the `example` folder.
native = "example/7d6y_1_B.pdb"
pred = "example/pred_7d6y_1_B.pdb"
rmsd = calc_ab_rmsd(native,pred)
print(rmsd)

"""
output:
{'CDRH1': tensor(0.3136), 'CDRH2': tensor(0.0898), 'CDRH3': tensor(4.3704), 'fv-H': tensor(0.6426)}
"""

In the terminal:

abrmsd --pred pred_7d6y_1_B.pdb --native 7d6y_1_B.pdb --verbose

# Output:
[INFO] Renumbered chain B (H)
[INFO] Renumbered chain A (H)
    _    _       ____  __  __ ____  ____  
   / \  | |__   |  _ \|  \/  / ___||  _ \ 
  / _ \ | '_ \  | |_) | |\/| \___ \| | | |
 / ___ \| |_) | |  _ <| |  | |___) | |_| |
/_/   \_\_.__/  |_| \_\_|  |_|____/|____/ 

>>> Result
Frag    RMSD(Å)
CDRH1   0.3136
CDRH2   0.0898
CDRH3   4.3704
fv-H    0.6426
>>> End

Calculate the RMSD between paired antibody structures (containing heavy and light chains).

from ab_rmsd import calc_ab_rmsd

# two example pdb files are provided in the `exampl`e folder.
native = "example/7s0b_.pdb"
pred = "example/pred_7s0b_.pdb"
rmsd = calc_ab_rmsd(native,pred)
print(rmsd)

"""
output:
{
    'CDRH1': tensor(25.6173), 'CDRH2': tensor(15.5819), 'CDRH3': tensor(25.7562), 'fv-H': tensor(15.9964), 
    'CDRL1': tensor(11.8419), 'CDRL2': tensor(13.8057), 'CDRL3': tensor(17.1446), 'fv-L': tensor(15.9478)
}
"""

Calculate the DockQ scores between predicted and native complex structure.

from DockQ.DockQ import calc_DockQ

scores = calc_DockQ(model='example/pred_7s0b_.pdb',native='example/7s0b_.pdb')
print(scores)
"""
The first four scores are usually used to evaluate the docking performance.
{
    'DockQ': 0.011549873197136384, 
    'irms': 17.429353912635577,
    'Lrms': 50.73969606449461, 
    'fnat': 0.0,
    'nat_correct': 0, 'nat_total': 55, 'fnonnat': 1.0, 
    'nonnat_count': 9, 'model_total': 9, 
    'chain1': 'A', 'chain2': 'B', 'len1': 121, 
    'len2': 107, 'class1': 'receptor', 'class2': 'ligand'
}
"""

Use the cli to calculate the DockQ score.

./DockQ/DockQ.py example/pred_7s0b_.pdb example/7s0b_.pdb

TODOs

• add DockQ as an evaluation for heavy and light chain complex structure.
• for multi-chain antibody, superimpose each chain seperately and calculate the RMSD. instead of superimposing the complex structure.

Credits

• Part of the code is adapted from shitong's Diffab.
• Code about the DockQ is from https://github.com/bjornwallner/DockQ.
• Huge thanks to @JinyuanSun for contributing the cli tool.