A Quality Measure for Protein, Nucleic Acids and Small Molecule Docking Models
With pip:
pip install DockQ
Or, if you want the latest commit: clone the repository and run pip within the repo directory:
git clone https://github.com/bjornwallner/DockQ/
cd DockQ
pip install .
After installing DockQ with pip, the DockQ binary will be in your path. Just run DockQ with:
DockQ <model> <native>
Example
When running DockQ on model/native complexes with one or more interfaces, you will get a result for each interface. Results are computed to maximise the average DockQ across all interfaces:
$ DockQ examples/1A2K_r_l_b.model.pdb examples/1A2K_r_l_b.pdb
****************************************************************
* DockQ *
* Docking scoring for biomolecular models *
* DockQ score legend: *
* 0.00 <= DockQ < 0.23 - Incorrect *
* 0.23 <= DockQ < 0.49 - Acceptable quality *
* 0.49 <= DockQ < 0.80 - Medium quality *
* DockQ >= 0.80 - High quality *
* Ref: Mirabello and Wallner, 'DockQ v2: Improved automatic *
* quality measure for protein multimers, nucleic acids *
* and small molecules' *
* *
* For comments, please email: [email protected] *
****************************************************************
Model : examples/1A2K_r_l_b.model.pdb
Native : examples/1A2K_r_l_b.pdb
Total DockQ over 3 native interfaces: 0.653 with BAC:ABC model:native mapping
Native chains: A, B
Model chains: B, A
DockQ: 0.994
iRMSD: 0.000
LRMSD: 0.000
fnat: 0.983
fnonnat: 0.008
F1: 0.987
clashes: 0
Native chains: A, C
Model chains: B, C
DockQ: 0.511
iRMSD: 1.237
LRMSD: 6.864
fnat: 0.333
fnonnat: 0.000
F1: 0.500
clashes: 0
Native chains: B, C
Model chains: A, C
DockQ: 0.453
iRMSD: 2.104
LRMSD: 8.131
fnat: 0.500
fnonnat: 0.107
F1: 0.641
clashes: 0
A more compact output option is available with the flag --short:
$ DockQ examples/1A2K_r_l_b.model.pdb examples/1A2K_r_l_b.pdb --short
Total DockQ over 3 native interfaces: 0.653 with BAC:ABC model:native mapping
DockQ 0.994 iRMSD 0.000 LRMSD 0.000 fnat 0.983 fnonnat 0.008 F1 0.987 clashes 0 mapping BA:AB examples/1A2K_r_l_b.model.pdb B A -> examples/1A2K_r_l_b.pdb A B
DockQ 0.511 iRMSD 1.237 LRMSD 6.864 fnat 0.333 fnonnat 0.000 F1 0.500 clashes 0 mapping BC:AC examples/1A2K_r_l_b.model.pdb B C -> examples/1A2K_r_l_b.pdb A C
DockQ 0.453 iRMSD 2.104 LRMSD 8.131 fnat 0.500 fnonnat 0.107 F1 0.641 clashes 0 mapping AC:BC examples/1A2K_r_l_b.model.pdb A C -> examples/1A2K_r_l_b.pdb B C
A dictionary containing the complete results data can be dumped in JSON format with the --json filename.json flag:
$ DockQ examples/1A2K_r_l_b.model.pdb examples/1A2K_r_l_b.pdb --json filename.json
By default, DockQ will try to find the optimal mapping between interfaces found in the native and in the model.
The simplest case is when a homodimer has been modelled. Then, the native interface "AB" (between native chains A and B) could be compared to either the model AB interface, but also BA, as changing the order of the chains will generally change the results. If the user runs:
DockQ homodimer_model.pdb homodimer_native.pdb
the software will report the mapping (AB -> AB or AB -> BA) with highest DockQ score.
If the user wishes to enforce a certain mapping, the flag --mapping can be used. This is useful, for example, when model/native contain a large number of homologous chains,
as it will speed up computations.
Complete mapping
The user defines the complete mapping between native and model chains with: --mapping MODELCHAINS:NATIVECHAINS. For example, in the previous case, two possible mappings can be:
--mapping AB:AB(native chain A corresponds to model chain A, native B to model B),--mapping AB:BA(native chain A corresponds to model chain B, native B to model A).
The pair before the colon : defines the chain order in the model, the pair after defines the order in the native.
Partial mapping
If the user wishes to fix part of the mapping and let DockQ optimize the rest, wildcards can be used. For example, if a tetramer has chains ABCD in the model and WXYZ in the native,
the user might use:
--mapping A*:W*
where the wildcard * indicates that DockQ should optimize the mapping between BCD and XYZ while keeping A -> W fixed. Multiple chains can be fixed:
--mapping AD*:WY*
Limiting the search to subsets of native interfaces
If the user is interested one or more specific interfaces in the native, while the rest should be ignored, the following can be used:
--mapping :WX
which is equivalent to:
--mapping *:WX
Then DockQ will find the interface in the model that best matches the WX interface in the native. Multiple native interfaces can be included:
--mapping :WXY
--mapping *:WXY
Small molecules in PDB or mmCIF files can be scored and the mapping optimized in the same way as for proteins. Just add the flag --small_molecules:
# Compare docking of hemoglobin chains (chain A and B in native) as well as HEM and PO4 groups (chains E, F, G)
$ DockQ examples/1HHO_hem.cif examples/2HHB_hem.cif --small_molecule --mapping :ABEFG --short
Total DockQ-small_molecules over 4 native interfaces: 0.659 with ABDCF:ABEFG model:native mapping
DockQ 0.950 iRMSD 0.455 LRMSD 1.451 fnat 0.964 fnonnat 0.070 clashes 0.000 F1 0.946 DockQ_F1 0.945 mapping AB:AB examples/1HHO_hem.cif A B -> examples/2HHB_hem.cif A B
LRMSD 0.585 mapping AD:AE (HEM) examples/1HHO_hem.cif A D -> examples/2HHB_hem.cif A E
LRMSD 28.096 mapping BC:BF (PO4) examples/1HHO_hem.cif B C -> examples/2HHB_hem.cif B F
LRMSD 1.311 mapping BF:BG (HEM) examples/1HHO_hem.cif B F -> examples/2HHB_hem.cif B G
Only LRMSD is reported for small molecules.
NB: Small molecules must be in the PDB/mmCIF files. They also must have separate chain identifiers (the label_asym_id field is used in mmCIF formatted files).
Interfaces involving nucleic acids are seamlessly scored along with protein interfaces. The DockQ score is calculated for protein-NA or NA-NA interfaces in the same way as for protein-protein interfaces (two DockQ scores are reported for double helix chains).
Run DockQ with -h/--help to see a list of the available flags:
usage: DockQ [-h] [--capri_peptide] [--small_molecule] [--short] [--verbose] [--no_align] [--n_cpu CPU] [--max_chunk CHUNK] [--optDockQF1] [--allowed_mismatches ALLOWED_MISMATCHES] [--mapping MODELCHAINS:NATIVECHAINS] <model> <native>
DockQ - Quality measure for protein-protein docking models
positional arguments:
<model> Path to model file
<native> Path to native file
optional arguments:
-h, --help show this help message and exit
--capri_peptide use version for capri_peptide (DockQ cannot not be trusted for this setting)
--small_molecule If the docking pose of a small molecule should be evaluated
--short Short output
--verbose, -v Verbose output
--no_align Do not align native and model using sequence alignments, but use the numbering of residues instead
--n_cpu CPU Number of cores to use
--max_chunk CHUNK Maximum size of chunks given to the cores, actual chunksize is min(max_chunk,combos/cpus)
--optDockQF1 Optimize on DockQ_F1 instead of DockQ
--allowed_mismatches ALLOWED_MISMATCHES
Number of allowed mismatches when mapping model sequence to native sequence.
--mapping MODELCHAINS:NATIVECHAINS
Specify a chain mapping between model and native structure. If the native contains two chains "H" and "L" while the model contains two chains "A" and "B", and chain A is a model of native chain H and chain B is a model
of native chain L, the flag can be set as: '--mapping AB:HL'. This can also help limit the search to specific native interfaces. For example, if the native is a tetramer (ABCD) but the user is only interested in the
interface between chains B and C, the flag can be set as: '--mapping :BC' or the equivalent '--mapping *:BC'.
Importing DockQ as python module
Once DockQ is installed with pip, it can also be used as a module in your python code:
from DockQ.DockQ import load_PDB, run_on_all_native_interfaces
model = load_PDB("examples/1A2K_r_l_b.model.pdb")
native = load_PDB("examples/1A2K_r_l_b.pdb")
# model:native chain map dictionary for two interfaces
chain_map = {"A":"A", "B":"B"}
# returns a dictionary containing the results and the total DockQ score
run_on_all_native_interfaces(model, native, chain_map=chain_map)
({('A', 'B'): {'DockQ_F1': 0.9437927182141027,
'DockQ': 0.9425398964102757,
'iRMSD': 0.3753064373774967,
'LRMSD': 0.5535111803522507,
'fnat': 0.8907563025210085,
'nat_correct': 106,
'nat_total': 119,
'fnonnat': 0.1016949152542373,
'nonnat_count': 12,
'model_total': 118,
'clashes': 0,
'len1': 124,
'len2': 124,
'class1': 'ligand',
'class2': 'receptor',
'chain1': 'A',
'chain2': 'B',
'chain_map': {'A': 'A', 'B': 'B'}}},
0.9425398964102757)
Merge multiple chains in a single receptor or ligand
See issue #33 if you want to merge multiple chains (e.g. heavy and light antibody chains) into a single receptor or ligand
DockQ: DockQ score as defined in the original DockQ paperiRMSD: RMSD of interfacial residuesLRMSD: Ligand RMSDfnat: Fraction of retrieved native contacts (same as Recall or TPR)fnonnat: Fraction of predicted contacts that are not native (same as FPR)F1: Harmonic mean of the precision and recall in predicted interfacial contactsclashes: number of clashing interfacial residues in model (distance between residues below 2Å)
If you use DockQ v2, please cite the preprint: https://doi.org/10.1101/2024.05.28.596225