/protein-loop-reconstruction

This repository provides results and code for the project "Protein loop reconstruction by combination of MD modeling and crystallographic data"

Primary LanguagePython

Protein loop reconstruction by using MD simulations and crystallographic data

This repository provides results and code for the project "Protein loop reconstruction by using MD simulations and crystallographic data"

Author:

  • Polina Vaganova (Saint Petersburg State University, Saint Petersburg, Russia)

Supervisors:

  • Olga Lebedenko (BioNMR laboratory, Saint Petersburg State University, Saint Petersburg, Russia)
  • Nikolai Skrynnikov (BioNMR laboratory, Saint Petersburg State University, Saint Petersburg, Russia)

Table of contents:

Background:

To date, X-ray crystallography remains a primary experimental method for elucidation of three-dimensional (3D) protein structure. The diffraction process is fundamentally different from microscopic imaging; crystallography is not an imaging technique. The crucial difference is that visible light scattered from objects can be focused through refractive lenses to create a magnified image of the object. This is not the case for X-rays, which are also electromagnetic radiation but of several orders of magnitude shorter wavelength and correspondingly higher energy; the refractive index of X-rays in different materials si essentially equal and close to unity, and no refractive lenses can be constructed for X-rays. Instead, the electron density of the scattering molecular structure must be reconstructed by special technique (Fourier transform). In itself, the Fourier reconstruction from reciprocal diffraction space back into direct molecular space poses no difficulties in principle, with the unfortunate qualifier that for this type of reconstruction two terms are needed as Fourier coefficients: the structure factor amplitudes, readily accessible in the form of the square root of the measured ($F_{obs}$) and corrected ($F_{calc}$) diffraction spot intensities (Figure 1); and as a second term for each observed diffraction spot, its relative phase angle.

xray-experiment.png
Figure 1. The principle of X-ray structure determination.

These phase angles are not directly accessible and must be adjusted during special procedure, called refinement. In refinement procedure we are adjusting phase angles to obtain a best fit between the observed structure factor amplitudes (($F_{obs}$)) and the computed model structure factor amplitudes ($F_{calc}$). The overall fit between diffraction data and model is numerically quantified by a global linear residual (the R-value) between the scaled structure factor amplitudes $F_{obs}$ and $F_{calc}$: $R = \frac{{\sum\limits_{hkl} |F_{\text{obs}} - F_{\text{calc}}|}}{{\sum\limits_{hkl} |F_{\text{obs}}|}}$, where $hkl$ Miller indexes for each of reflection spots.

The X-ray crystallography technique operates with static images of molecular structures. The 3D structure is formed by secondary struture elements (e.g. α-helices and β-strands) as well as loosely structured elements such as loops. Protein loops are often functionally important. Their conformational plasticity is key to molecular recognition, allosteric control, ligand binding, and signaling.

However, the structural variability of protein loops presents a challenge for the X-ray crystallography which is normally limited to static structural models. As a consequence, mobile loops are often absent from the crystallographic structures deposited in the Protein Data Bank (PDB). Nevertheless, residual electron density associated with such loops can be used to rebuild them in a form of conformational ensemble. To this end, we have used Molecular Dynamics (MD) simulations additionally guided by the experimental diffraction data. The appropriate crystallographic refinement procedure has been developed in our laboratory using the biomolecular MD simulation platform Amber [1].

Goal: The main goal of this project is to reconstruct the protein loop ensemble using Molecular Dynamics (MD) simulation guided by the experimental diffraction data.

Methods:

Protein structures with missing loop regions and different levels of residual electron density at the disordered loop sites have been identified by the automated parsing of RCSB database. An example of one such structure (PDB: 1k33) is illustrated in Figure 2.

$R_{work}$ / $R_{free}$ for this deposited structure are 0.206 / 0.226.

Initial loop conformations were generated using programs such as Modeller or Rapper (Figure 3). This step doesn't operate with electron density data and thus the resulting structure has worse scores (higher is worse).

$R_{work}$ / $R_{free}$ for Modeller resulting structure are 0.316 / 0.317.

Subsequently, MD-based crystallographic refinement procedure has been performed on structural models in a form of crystal unit cells or supercells (Figure 4).

The refined models involving multiple loop conformations (Figure 5) have been validated against the available electron density maps using phenix.molprobity module.

1k33_without_loop.png 1k33_with_loop.png 1k33_ucell.png 1k33_loop_ensemble.png
Figure 2. Structure of 1k33 without loop residues. Figure 3. Structure of 1k33 with built loop residues (red). Figure 4. Structure of 1k33 unit cell. Figure 5. Loop ensemble of 1k33.

After refinement procedure we observed the dramatic improvement in $R_{work}$ / $R_{free}$ from 0.206 / 0.226 (RCSB) to 0.181 / 0.202 in our new approach.

System requirements:

Key packages and programs:

  • Python (>= 3.9)
  • Amber22 (a build with MPI is employed)
  • Modeller (10.4) package for loop modelling
  • phenix software package for macromolecular structure determination
  • pyxmolpp2 (1.6.0) in-house python library for processing molecular structures and MD trajectories
  • slurm (20.11.8) cluster management and job scheduling system
  • other python libraries used are listed in requirements.txt

Repository structure:

  • 0_prepare_annotation contains rcsb annotation results
  • 1_annotated_rcsb contains annotation only for monomers with different symmetry operations
  • 2_rcsb_data contains examples for input files
  • 3_loop_building contains examples for building initial loop conformation
  • 4_protocol_run contains examples for running MD-based refinement protocol
  • arx contains special module for AMBER
  • utils contains python scripts for annotations, filtering, loop building and input preparation
  • figures contains figures for README.md you are currently reading

Single structure loop building: example of 1k33

  1. Download a PDB file (coordinates), CIF file (structure factors) and FASTA file (aminoacid sequence) from the RCSB database. Place files into corresponding folders in 2_rcsb_data dir.

  2. Go to 2_rcsb_data/mtz dir and convert CIF to MTZ using Phenix. Then expand it to P1 space group:

cd 2_rcsb_data/mtz
phenix.cif_as_mtz ../cif/1k33-sf.cif --ignore_bad_sigmas --merge --output_file_name=1k33-sf.mtz
phenix.reflection_file_converter --expand_to_p1 1k33-sf.mtz --write_mtz_amplitudes --mtz_root_label="FOBS" --label="FOBS" --generate_r_free_flags --non_anomalous --mtz 1k33.mtz
  1. Build initial loop using Modeller. Please, adjust the pdb_ids variable in the run_phenix.sh accordingly before running it.
cd ../..
python utils/modeller/model.py

  1. Create topology (and coordinate) file from PDB:

    • Copy pdn and P1-MTZ files into input dir:
    python utils/prepare_inputs/1_copy_inputs.py

    • source AMBER, since we rely on its Python libraries.

    • Set up Python virtual environment and install dependencies (to avoid potential conflicts, use amber.python):

    amber.python -m venv venv
source venv/bin/activate
pip install -U pip wheel setuptools
pip install -r requirements.txt
    • Add the repo's directory into PYTHONPATH variable
    export PYTHONPATH=$PYTHONPATH:$(pwd)
    • Run the script to prepare a model
    python utils/prepare_inputs/2_prepare-structures.py

    The results will be written to 4_protocol_run/amber-topology/1k33/.

  2. Prepare the refinement job:

python init.py

First, as a preparatory step for the x-ray restrained MD run, one needs to convert MTZ into a simple text format ( described in the AMBER manual). File conversion is performed by write_sf_dat_file() method of init.py. Second, one needs to create x-ray-specific topology file. The expansion of the topology file (adding crystallographic parameters) is performed by prepare_xray_prmtop() method of init.py. The files are written to 4_protocol_run/output/1k33/. Note that the script automatically reads the number of residues from the topology file and uses it during the refinement (heating, evolution and cooling stages).

If you wish to employ a 2x2x2 supercell model instead of the unit cell model, please, open the file init.py and add .sc postfix to the name of the folder in the main() function (i.e. replace prepared with prepared.sc).

  1. Finally, run the refinement job locally by executing:
python run_locally.py

Alternatively, if you want to use a different machine, you should employ LocalSlurmWorker remote runner and execute python run_remotely.py instead. NB: Don't forget to adjust paths that are sourced and exported (in essence, environmental variables should be defined similar to step 3). The results will be written to 4_protocol_run/output/1k33.

  1. Water picking, B-factors refinement and MolProbity reports generation

Historically, these tasks were executed separately from the AMBER-based pipeline with phenix software [2]. The related files can be found in the 4_protocol_run/phenix_refinment directory. Please, adjust the paths in the run_phenix_remotely.py accordingly before running it (change {your_path_to_repo} with your actually path).

If you want to run locally without slurm then launch residue_renamer.py on your pdb, adjust the paths in the phenix.sh and launch it. The first argument of the script should be the path to the PDB file, the second one - the path to the MTZ file.

References

[1] Mikhailovskii, O., Xue, Y., Skrynnikov, N. R. Modeling a Unit Cell: Crystallographic Refinement Procedure Using the Biomolecular MD Simulation Platform Amber. 2022. IUCrJ, 9 (1): 114–133. https://doi.org/10.1107/S2052252521011891.

[2] Liebschner, D., Afonine, P. V., Baker, M. L. et al. 2019. Macromolecular structure determination using x-rays, neutrons and electrons: recent developments in phenix. Acta Crystallogr D Struct Biol. 75, 861–877. https://doi.org/10.1107/S2059798319011471