Nanobody - Antigen HADDOCK docking

This repository contains all data and analysis required to reproduce the figures and supplementary figures in the paper 'Combining AI structure prediction and integrative modelling for nanobody-antigen complexes' from Miguel Sánchez-Marín, Marco Giulini, Alexandre MJJ Bonvin

Repository organisation

This repository contains the follwoing directories:

• analysis: Python scripts required to reproduce the paper figures.

  • figures/: directory with the report figures in png format.
  • af_metrics_vs_sr.py : reads the success rates and metrics for both Alphafold2-Multimer and AlphaFold3 and compares them to the docking success rates, highlighting the cases when HADDOCK docking was able to "rescue" a failed AlphaFold run.
  • dockq_features_correlation.py : reads the files data/*_dockq.tsv and performs several analysis with the correlation between the max dockQ in the top10 models and several quantities, such as the minimum CDR3 RMSD across the ensemble (see ).
  • flexref_fnat_improvement.py : reads the file data/fnat_diff_flexref.tsv and plots the Fnat improvement distribution after flexible refinement stage (analysis/figures/figure4.png).
  • haddock_docking_sr_scenario_analysis.py : reads the files data/haddock_docking_sr_per_scenario.tsv, and data/haddock_docking_clustered_sr_per_scenario.tsv and plots the docking success rates.
  • paratope_analysis.py : reads the file data/probability_regions_interaction.tsvand plots the regions frequency in the paratope of kinked and extended nanobodies (figures/SI_figure5.png). Reads the file data/clustered_paratope_interacting_residues.tsv and plots the frequency of each residue in the different CDR and CDR+FR paratope representations (analysis/figures/figure5.png). Finally, it calculates and plots the docking success rates for these new restraints (analysis/figures/mixed_paratope_restraints_sr.png).
  • single_structure_assessment.py: does all the analysis about the monomeric nanobody structure, including calculating the mean, standard deviation and median values for all regions RMSD (analysis/figures/nanobody_rmsd_summary.tsv), plotting the CDR3 RMSD distribution for top1 models (analysis/figures/cdr3_rmsd_top1_prediction_comparison.png) and for the best in each ensemble (analysis/figures/cdr3_rmsd_best_ensemble_best_prediction_comparison.png). The script calculates also the RMSD distribution for the unbound antigens (analysis/figures/rmsd_unbound_antigen_comparison.png).

• benchmark_pdb_files: directory with all the input nanobody structures, ligand structures and restraint files used in the report. Quality metrics of each obtained HADDOCK model are also reported.

  • {pdb}{chain}: there is a directory for each PDB in the dataset containing:
  • *ensemble.pdb: the input nanobody ensemble from IB (ImmuneBuilder), IBMo (ImmuneBuilder + AlphaFold2), IBMu (ImmuneBuilder + AlphaFold2-Multimer) or IBMM (ImmuneBuilder + ALphaFold2 + AlphaFold2-Multimer) predictions.
  • {pdb}{chain}_allsurface_ambig.tbl: input ambiguous restraints for All Surface scenario.
  • {pdb}{chain}_l_u_unambig.tbl: input unambiguous restraints for unbound ligand structure.
  • {pdb}{chain}_real_ambig.tbl: input ambiguous restraints for Real Interface scenario.
  • {pdb}{chain}_l_b.pdb: input bound antigen structure.
  • {pdb}{chain}_loose_ambig.tbl: input ambiguous restraints for Loose Interface scenario.
  • {pdb}{chain}_ref.pdb: input reference structure.
  • {pdb}{chain}_l_b_unambig.tbl: input unambiguous restraints for bound ligand structure.
  • {pdb}{chain}_loose_mix_ambig.tbl.tgz: input ambiguous restraints for Mixed paratope Loose Interface scenario.
  • {pdb}{chain}_twohit_ambig.tbl: input ambiguous restraints for Two-Hit Interface scenario.
  • {pdb}{chain}_l_u.pdb: input unbound antigen structure.
  • {pdb}{chain}_r_b.pdb: bound nanobody structure.
  • {pdb}{chain}_twohit_mix_ambig.tbl.tgz: input ambiguous restraints for Mixed paratope Two-Hit Interface scenario.
  • {pdb}{chain}_capri.tsv: CAPRI quality metrics for all the generated HADDOCK models.

• cfg_files: example .cfg files for docking with HADDOCK3 for all different information scenarios.

  • nanobody_antigen_all_surface_haddock_docking.cfg: example .cfg file for All Surface docking.
  • nanobody_antigen_loose_interface_haddock_docking.cfg: example .cfg file for Loose Interface docking.
  • nanobody_antigen_mix_loose_interface_haddock_docking.cfg: example .cfg file for Mixed Loose Interface docking.
  • nanobody_antigen_mix_twohit_interface_haddock_docking.cfg: example .cfg file for Mixed Two-Hit Interface docking.
  • nanobody_antigen_real_interface_haddock_docking.cfg: example .cfg file for Real Interface docking.
  • nanobody_antigen_twohit_interface_haddock_docking.cfg: example .cfg file for Two-Hit Interface docking.

• data: all data required to reproduce the figures in the report.

  • *_rmsd.tsv: RMSD of the different nanbody regions (FR, CDR1, CDR2, CDR3) for the different methods. If ending with _centroids_rmsd.tsv it only represents CDR3 RMSD from the centroids of the clustered ensembles of predictions.
  • alphafold2* and alphafold3* : all the relevant statistics about AlphaFold2-Multimer and AlphaFold3 models.
  • cdr3_rmsd_dockq.tsv: minimum available CDR3 RMSD in the ensemble and maximum DockQ achieved in the top10 models
  • clustered_paratope_interacting_residues.tsv: paratope residues from each Paratope Dataset PDB and the cluster to which they belong.
  • epitope_rmsd_dockq.tsv minimum epitope RMSD and maximum DockQ achieved in the top10 models
  • fnat_diff_flexref.tsv: frequency of fnat improvements after flexible refinement classified in 0.05 width bins.
  • haddock_docking_*.tsv: docking success rates for different CAPRI classes and depending of topN ranked models/clusters, for different information scenarios and/or pipeline stages.
  • paratope_represented_dockq.tsv : % of represented paratope in the restraints and maximum DockQ achieved in the top10 models for Loose and Two-Hit Interface scenario, bound antigen EMRef stage.
  • probability_regions_interaction.tsv : percent of kinked/extended structures with each nanobody region being present in the paratope.
  • voro_scoring_*_sr.tsv: docking success rates for different information scenarios as a function of the topN models/clusters after VoroIF-GNN scoring the final HADDOCK-scored models.