DODO is a Python package and command-line utility for taking an AF2 structure and redesigning the disordered regions to make them look more like IDRs. To be clear, the work done by DeepMind to make AlphaFold2 is AMAZING, and I do not mean to take away from that in ANY WAY. However, for visualizing proteins for presentations, etc. it would be nifty to be able to make the IDRs look more 'IDR-like'. DODO does just that! What it does is identify the IDRs in the structure, predict the end-to-end distance for each IDR from its sequence using ALBATROSS (see https://github.com/idptools/sparrow) and rebuild the structure such that the IDRs are the approximate correct overall dimensions (see example below). If that visualization doesn't work for you, there are other options to make the IDRs more compact or expanded than is predicted from sequence. In addition, you can make a PDB with multiple IDRs in a single 'structure' and keep the folded domains fixed, which when opened in VMD makes something that looks like a simulation trajectroy (to be very clear, it is NOT the equivalent to an actual simulation trajectory but is really nice for visualizations).
-
Rebuilding employs a simple random walk approach, so the actual IDR conformation is completely random and not scientifically useful. However, it is still quite useful for visualization of the protein overall.
-
At this moment, the rebuilt IDRs only contain alpha carbons whereas folded regions contain all atoms. This is mainly to do with the trickiness of placing all the atoms back after dramatically changing the IDR. I'm working on all-atom IDR generation, which I'm hoping to bring in the future (but this is not a gaurantee because it's tricky business and for visualization is not really necessary to be honest).
-
Because some IDRs only have alpha carbons, the order of bonds result in an 'unusual bond' warning in VMD.
-
Some visualization modes don't work in VMD. Licorice seems to work fine, tube and trace do not. I'm working on this.
DODO is currently usable in Python and from the command-line.
If you are having any issues installing DODO, we have a few known potential install issues, so please keep reading this section! If you're having a problem that we don't have listed here, please let me know!
Note - to install DODO, you first need to have cython and numpy installed. To install cython and numpy, simpy run:
pip install cython numpyOnce you have cython and numpy installed, you should be able to install DODO.
To install DODO, run the following command from terminal:
pip install git+https://github.com/idptools/dodo.gitOver time we have discovered some ways that installing DODO can fail. One is related to your version of setuptools. Try running the following and then attempt to install DODO:
pip install setuptools --upgradeAnother fail that we are aware of involves wheel. To fix this, run the following:
pip install wheel --upgradeFirst import build from DODO.
from dodo import buildYou can build new structures from from an existing PDB or just have DODO download the structure from the AF2 database. You can also generate PDBs of IDRs from sequence alone!
To have DODO download a structure from a protein name and alter the disordered regions, you can use the pdb_from_name() function. There are two required arguments unless you set graph=True: 1. the protein name as a string, 2. the out_path for where to save the PDB. If you set graph=True, you don't need to specify the outpath and DODO will just show your PDB in a 3D graph using matplotlib.
build.pdb_from_name('human p53', out_path='/Users/your_user_name/Desktop/my_cool_proteins/my_protein.pdb')Additional usage:
All arguments for build.pdb_from_name() are as follows:
protein_name - required. The name of your protein as a string. Specifying the organism increases your chance of success.
out_path - optional if you set graph=True. Otherwise raises an exception. Where to save your protein structure file. Specify the file name here.
mode - optional. Default: 'predicted'. The predicted option predicts the end-to-end distance of your disordered regions from sequence and then makes the IDRs fit within that distance. Additional options are super_compact, compact, normal, expanded, super_expanded, max_expansion. These are pretty self explanatory.
num_models - optional. Default: 1. num_models lets you choose the number of models of IDRs to make for your protein. The folded domains are left in the same location for all models wherease the IDRs vary.
linear_placement - optional. Default: False. Whether to place the folded domains linearly for visualization.
just_fds - optional. Default: False. Setting just_fds to True will save out folded domains as individual PDBs with the name of your protein as specified in out_path with the coordinates of the fd in the file name. Formatted as protname_resStart_resEnd.pdb for each FD.
beta_for_FD_IDR - optional. Default: False. Whether to set beta values such that all IDRs = 0 and FDs=100 for visualization.
include_FD_atoms - optional. Default: True. Whether to include all atoms for the FDs. Only CA for IDRs for now.
CONECT_lines - optional. Default: True. Whether to included CONECT lines in the generated PDB. Makes visualization generally better.
verbose - optional. Default: True. Whether to show progress as structure is being made.
use_metapredict - optional. Default: False. This option lets you use metapredict to predicte the IDRs and folded regions. Although fairly accurate, it doesn't get the exact cutoffs for some regions and fails to predict small loops within large folded regions. The default is to use the number of atoms neighboring each atom in the AF2 structure. The default behavior is slower but works better.
graph - optional. Default: False. Setting this to True will pull up a really rough looking structure of your protein using the 3D graphing functionality in matplotlib. This is something I made when developing this to quickly look at structures. You shouldn't use this, but you can if you want. It's kind of fun TBH.
attempts_per_region - optional. Default: 20. Number of times to try and make each region.
attempts_per_coord - optional. Default: 2000. Number of times to try to generate each coordinate for each alpha carbon in the structure.
You can also have DODO alter a pre-existing AF2 pdb file. The AF2 file should have all atom information (though this isn't required). There are two required arguments if not graphing: path_to_pdb : the path to your pdb file as a string and out_path : the path and filename of where to save your file. If you set graph=True, you don't need to specify the out_path.
build.pdb_from_pdb('/Users/your_user_name/Desktop/my_AF2_pdb.pdb', out_path='/Users/your_user_name/Desktop/my_AF2_PDB_DODO.pdb')Additional usage:
All arguments for build.pdb_from_pdb() are as follows:
path_to_pdb - required. The filepath to your pdb as a string.
out_path - optional if you set graph=True. Otherwise raises an exception. Where to save your protein structure file. Specify the file name here.
mode - optional. Default: 'predicted'. The predicted option predicts the end-to-end distance of your disordered regions from sequence and then makes the IDRs fit within that distance. Additional options are super_compact, compact, normal, expanded, super_expanded, max_expansion. These are pretty self explanatory.
num_models - optional. Default: 1. num_models lets you choose the number of models of IDRs to make for your protein. The folded domains are left in the same location for all models wherease the IDRs vary.
linear_placement - optional. Default: False. Whether to place the folded domains linearly for visualization.
just_fds - optional. Default: False. Setting just_fds to True will save out folded domains as individual PDBs with the name of your protein as specified in out_path with the coordinates of the fd in the file name. Formatted as protname_resStart_resEnd.pdb for each FD.
beta_for_FD_IDR - optional. Default: False. Whether to set beta values such that all IDRs = 0 and FDs=100 for visualization.
include_FD_atoms - optional. Default: True. Whether to include all atoms for the FDs. Only CA for IDRs for now.
CONECT_lines - optional. Default: True. Whether to included CONECT lines in the generated PDB. Makes visualization generally better.
verbose - optional. Default: True. Whether to show progress as structure is being made.
use_metapredict - optional. Default: False. This option lets you use metapredict to predicte the IDRs and folded regions. Although fairly accurate, it doesn't get the exact cutoffs for some regions and fails to predict small loops within large folded regions. The default is to use the number of atoms neighboring each atom in the AF2 structure. The default behavior is slower but works better.
graph - optional. Default: False. Setting this to True will pull up a really rough looking structure of your protein using the 3D graphing functionality in matplotlib. This is something I made when developing this to quickly look at structures. You shouldn't use this, but you can if you want. It's kind of fun TBH.
attempts_per_region - optional. Default: 20. Number of times to try and make each region.
attempts_per_coord - optional. Default: 2000. Number of times to try to generate each coordinate for each alpha carbon in the structure.
If you just have an IDR sequence, you can also generate coordinates for that. Just like before, you can save it or you can graph it in matplotlib.
build.pdb_from_sequence('VQQQGIYNNGTIAVANQVSCQSPNQ', out_path='/Users/your_user_name/Desktop/my_AF2_PDB_DODO.pdb')Additional usage:
All arguments for build.pdb_from_sequence() are as follows:
sequence - The sequence of the IDR as a string.
out_path - optional if you set graph=True. Otherwise raises an exception. Where to save your protein structure file. Specify the file name here.
mode - optional. Default: 'predicted'. The predicted option predicts the end-to-end distance of your disordered regions from sequence and then makes the IDRs fit within that distance. For IDRs between folded domains, the folded domains are moved apart from each other to accomodate the IDR length. For terminal IDRs, they will just adopt a configuration within the distance equal to the end to end distance. Additional options are super_compact, compact, normal, expanded, super_expanded, max_expansion. These are pretty self explanatory.
CONECT_LINES - optional. Default: True. Writes CONECT lines to the PDB so that all alpha carbons are connected.
graph - optional. Default: False. Setting this to True will pull up a really rough looking structure of your protein using the 3D graphing functionality in matplotlib. This is something I made when developing this to quickly look at structures. You shouldn't use this, but you can if you want. It's kind of fun TBH.
verbose optional. Default: False. Set to True to get more info on what is happening as your structure is being made.
attempts_per_idr - optional. Default: 50. Number of attempts to IDR. Basically if number of attempts per coordinate reaches max and it still hasn't succeeded, this is the number of times it tries again.
attempts_per_res - optional. Number of attempts to make each coordinate fit without clashing.
end_coord - optional. Default: (0,0,0). The end coordinate for the IDR as a tuple.
Just like in Python, you can generate AF2 structures with re-designed IDRs from the command-line using an existing AF2 PDB file or the name of a protein, and you can also generate PDBs for coordinates of IDRs from sequence alone. The only thing missing from the command-line is the graphing functionality.
To have DODO download a structure using a protein name and alter the disordered regions, you can use the pdb-from-name command. There are two required arguments: 1. the protein name, 2. the out_path for where to save the PDB.
pdb-from-name human p53 -o /Users/your_user_name/Desktop/my_cool_proteins/my_protein.pdbAdditional usage:
All arguments for pdb-from-name are as follows:
-o or --out_path : required. Where to save your protein structure file. Specify the file name here.
-m or --mode : optional. Default: predicted. --mode lets you specify how to build the IDR. The default predicted option predicts the end-to-end distance of your disordered regions from sequence and then makes the IDRs fit within that distance. Additional options are super_compact, compact, normal, expanded, super_expanded, max_expansion. These are pretty self explanatory.
-n or --num_models : optional. Default: 1. --num_models lets you choose the number of models of IDRs to make for your protein. The folded domains are left in the same location for all models wherease the IDRs vary.
-l or --linear_placement : optional. Default: False. The --linear_placement flag lets you place the IDRs linearly across a vector for visualization.
-j or --just_fds : optional. Default: False. The --just_fds flag will save out folded domains as individual PDBs with the name of your protein as specified in out_path with the coordinates of the fd in the file name. Formatted as protname_resStart_resEnd.pdb for each FD.
-b or --beta_for_FD_IDR : optional. Default: False. Whether to set beta values such that all IDRs = 0 and FDs=100 for visualization.
-c or --no_CONECT_lines : optional. Default: False. The --no_CONECT_lines flag lets you make files without CONECT lines.
-f or --no_FD_atoms : optional. Default: False. The --no_FD_atoms flag lets you make structures with ONLY alpha carbon atoms. By default, the IDRs are only alpha carbons and the FDs are all atom.
-u or --use_metapredict : optional. Default: False. The --use_metapredict flag lets use metapredict V2-ff to predict the disordered regions in your structure. By default, the location of all atoms in the structure are used to infer the location of the IDRs.
-s or --silent : optional. Default: False. The --silent flag lets you silent most print output to your terminal.
-apr or --attempts_per_region : optional. Default: 40. --attempts_per_region lets you specify the number of attempts to make each region of the structure.
-apc or --attempts_per_coord : optional. Default: 2000. --attempts_per_coord lets you specify the number of attempts to make to generate each coordinate in your structure.
To have DODO redesign the disordered regions using an AF2 PDB you have saved locally, you can use the pdb-from-pdb command. There are two required arguments: 1. The path to your PDB file including the filename and extension and, 2. the out_path for where to save the final PDB.
pdb-from-pdb /Users/your_user_name/Desktop/my_AF2_proteins/AF2_protein.pdb -o /Users/your_user_name/Desktop/my_cool_proteins/my_protein_with_a_new_IDR.pdbAdditional usage:
All arguments for pdb-from-pdb are as follows:
-o or --out_path : required. Where to save your protein structure file. Specify the file name here.
-m or --mode : optional. Default: predicted. --mode lets you specify how to build the IDR. The default predicted option predicts the end-to-end distance of your disordered regions from sequence and then makes the IDRs fit within that distance. Additional options are super_compact, compact, normal, expanded, super_expanded, max_expansion. These are pretty self explanatory.
-n or --num_models : optional. Default: 1. --num_models lets you choose the number of models of IDRs to make for your protein. The folded domains are left in the same location for all models wherease the IDRs vary.
-l or --linear_placement : optional. Default: False. The --linear_placement flag lets you place the IDRs linearly across a vector for visualization.
-j or --just_fds : optional. Default: False. The --just_fds flag will save out folded domains as individual PDBs with the name of your protein as specified in out_path with the coordinates of the fd in the file name. Formatted as protname_resStart_resEnd.pdb for each FD.
-b or --beta_for_FD_IDR : optional. Default: False. Whether to set beta values such that all IDRs = 0 and FDs=100 for visualization.
-c or --no_CONECT_lines : optional. Default: False. The --no_CONECT_lines flag lets you make files without CONECT lines.
-f or --no_FD_atoms : optional. Default: False. The --no_FD_atoms flag lets you make structures with ONLY alpha carbon atoms. By default, the IDRs are only alpha carbons and the FDs are all atom.
-u or --use_metapredict : optional. Default: False. The --use_metapredict flag lets use metapredict V2-ff to predict the disordered regions in your structure. By default, the location of all atoms in the structure are used to infer the location of the IDRs.
-s or --silent : optional. Default: False. The --silent flag lets you silent most print output to your terminal.
-apr or --attempts_per_region : optional. Default: 40. --attempts_per_region lets you specify the number of attempts to make each region of the structure.
-apc or --attempts_per_coord : optional. Default: 2000. --attempts_per_coord lets you specify the number of attempts to make to generate each coordinate in your structure.
If you just have an IDR sequence, you can also generate coordinates for that.
pdb-from-sequence GRNQNGGGYQNYNNQGYQGHGGQHQNNYNQYPCNYFGPGYNN -o /Users/your_user_name/Desktop/my_cool_proteins/my_cool_IDR.pdbAdditional usage:
All arguments for pdb-from-pdb are as follows:
-o or --out_path : required. Where to save your protein structure file. Specify the file name here.
-m or --mode : optional. Default: predicted. --mode lets you specify how to build the IDR. The default predicted option predicts the end-to-end distance of your disordered regions from sequence and then makes the IDRs fit within that distance. Additional options are super_compact, compact, normal, expanded, super_expanded, max_expansion. These are pretty self explanatory.
-n or --num_models : optional. Default: 1. --num_models lets you choose the number of models of IDRs to make for your protein. The folded domains are left in the same location for all models wherease the IDRs vary.
-c or --no_CONECT_lines : optional. Default: False. The --no_CONECT_lines flag lets you make files without CONECT lines.
-api or --attempts_per_IDR : optional. Default: 50. --attempts_per_IDR lets you specify the number of attempts to make to the IDR.
-apr or --attempts_per_residue : optional. Default: 1000. --attempts_per_residue lets you specify the number of attempts to make the coordinates for each residue for your IDR.
Logging changes below.
-
V0.15 - July 9, 2024. Fixed bug where you couldn't make structures from local PDBs when using metapredict if the structure was predicted to be completely disordered. Also fixed the inability to have multiple models for local PDBs that are fully disordered.
-
V0.14 - April 9, 2024. Fixed bug where you couldn't manually assign regions in the pdb_from_pdb function. Still need to add user-facing docs for this!
-
V0.13 - December 11, 2023. Fixed bug where proteins predicted to be fully disordered failed to build. Shoutout to Github user alexpmagalhaes for pointing out the bug and reporting with enough information to make it a quick fix.
-
V0.12 - November 15, 2023. Added ability to save out the predicted folded domains as individual PDBs from Python and in the command line.
-
V0.11 - October 30, 2023. Couple small fixes, updated documentation, fixed some embarassingly bad typos.
-
V0.10 - October 24, 2023. Big changes! Added functionality to generate multiple IDRs for a single PDB so you can make 'simulation-like' visualizations when viewing in PDB! I also added command-line functionality and fixed some more bugs. Note: the multiple models are for visualization only and not equivalent to actual simulations!
-
V0.06 - October 17, 2023. Added functionality to place folded domains in an approximate linear arrangement for visualization purposes.
-
V0.05 - October 5, 2023. Major overhaul to the backend to make structure generation more robust and efficient. Changed some user facing functionality and added ability to generate a PDB of an IDR from sequence alone. Improved documentation.
-
V0.04 - September 29, 2023. Made it so the atoms of folded domains can be kept in the structure. Only CA for the IDRs for now. Improved performance a bit. Still need to clean up the code a lot.. Made everything callable by the PDBParserObj made from the class pdb_tools.PDBParser
-
V0.03 - September 28, 2023. Made it so you can save the IDRs generated from sequence alone. Removed the ability to make all atom structures because it wasn't great. This hopefully will come in the future.
-
V0.02 - September 26, 2023. Added generating IDR coords from sequence alone. Added filling in all atom coordinates from alpha carbon coordinate using fixed bond angles and distances, which isn't great but is better than nothing.
-
V0.01 - September 25, 2023. Initial release.
Copyright (c) 2023, Ryan Emenecker - Holehouse Lab
Project based on the Computational Molecular Science Python Cookiecutter version 1.1.