Protein structure predictions for Pfam families
===============================================

This directory provides structural models and contact maps 
for 9648 Pfam families (Pfam 35.0).

Models were generated by the Baker group (https://www.bakerlab.org/) 
using RoseTTAfold (1) from multiple sequence alignments. 


Overview
--------

There are two archives available for download:
    * RoseTTAfold.full.tar.gz - the complete data, which includes  
      contact maps, pdb files, lDDT scores and alignments

    * RoseTTAfold.pdb.tar.gz - a smaller archive containing
      only the pdb files

    Note that these RoseTTAfold archives contain some additional 
    low quality models that are not shown on the InterPro and Pfam 
    websites.

Download the archive and verify its integrity:

    Full archive:
    $ wget http://ftp.ebi.ac.uk/pub/databases/Pfam/releases/Pfam35.0/structure_models/RoseTTAfold.full.tar.gz
    $ wget http://ftp.ebi.ac.uk/pub/databases/Pfam/releases/Pfam35.0/structure_models/RoseTTAfold.full.tar.gz.md5
    $ md5sum -c RoseTTAfold.full.tar.gz.md5

    Smaller pdb archive:
    $ wget http://ftp.ebi.ac.uk/pub/databases/Pfam/releases/Pfam35.0/structure_models/RoseTTAfold.pdb.tar.gz
    $ wget http://ftp.ebi.ac.uk/pub/databases/Pfam/releases/Pfam35.0/structure_models/RoseTTAfold.pdb.tar.gz.md5
    $ md5sum -c RoseTTAfold.pdb.tar.gz.md5

    

Untar the archive:

    Full archive:
    $ tar -xzf RoseTTAfold.full.tar.gz
    $ ls -l 
        drwxr-xr-x 5 interpro interpro        4096 Feb 25 22:49 RoseTTAfold_full
        -rw-r--r-- 1 interpro interpro 87255064553 Feb 25 22:49 RoseTTAfold.full.tar.gz
        -rw-r--r-- 1 interpro interpro          56 Feb 25 22:49 RoseTTAfold.full.tar.gz.md5

    Smaller pdb archive:
    $ tar -xzf RoseTTAfold.pdb.tar.gz
    $ ls -l 
        drwxr-xr-x 5 interpro interpro      512000 Feb 25 22:49 pdb
        -rw-r--r-- 1 interpro interpro   321154409 Feb 25 22:49 RoseTTAfold.pdb.tar.gz
        -rw-r--r-- 1 interpro interpro          55 Feb 25 22:49 RoseTTAfold.pdb.tar.gz.md5


If you downloaded the full archive, the RoseTTAfold_full directory has the following structure:

    .
    |-- a3m
    |   |-- PF00242.a3m
    |   |-- PF00257.a3m
    |   |-- ...
    |   `-- PF20139.a3m
    |-- npz
    |   |-- PF00242.npz
    |   |-- PF00257.npz
    |   |-- ...
    |   `-- PF20139.npz
    `-- pdb
        |-- PF00242.pdb
        |-- PF00257.pdb
        |-- ...
        `-- PF20139.pdb


    * a3m: contains the UniProt alignments for Pfam families, in the A3M file format.
    
    * npz: contains the contact maps with a pre-set distance threshold of 15Å, the predicted 
           residue-wise local Distance Difference Test (lDDT) scores (2) and the distance 
           error distributions, all in the NumPy NPZ file format.

    * pdb: directory of predicted structures, in the PDB file format.

If you downloaded the smaller RoseTTAfold.pdb.tar.gz archive, you will only have the pdb directory
containing the PDB files.


Contact maps
------------

Contact maps can be retrieved from the files in the /npz/ directory, 
using the Python code below (requires the NumPy package).

    $ python
    Python 3.6.9 (default, Oct  8 2020, 12:12:24) 
    [GCC 8.4.0] on linux
    Type "help", "copyright", "credits" or "license" for more information.
    >>> import numpy as np
    >>> data = np.load("RoseTTAfold_full/npz/PF00242.npz")
    >>> dist = data["mask"]

dist is a matrix of shape [L, L], where L is the size of the sequence.


lDDT scores
-----------

The local Distance Difference Test (lDDT) scores can be retrieved from the files
in the /npz/ directory.

    $ python
    Python 3.6.9 (default, Oct  8 2020, 12:12:24) 
    [GCC 8.4.0] on linux
    Type "help", "copyright", "credits" or "license" for more information.
    >>> import numpy as np
    >>> data = np.load("RoseTTAfold_full/npz/PF00242.npz")
    >>> lddt = data["lddt"]

lddt is an array of the size of the sequence.


Distance error distribution
---------------------------

The distance error distribution can be retrieved from the files
in the /npz/ directory.


    $ python
    Python 3.6.9 (default, Oct  8 2020, 12:12:24) 
    [GCC 8.4.0] on linux
    Type "help", "copyright", "credits" or "license" for more information.
    >>> import numpy as np
    >>> data = np.load("RoseTTAfold_full/npz/PF00242.npz")
    >>> distance_error = data["estogram"]

distance_error is a matrix of shape [L, L], where L is the size of the sequence.

References
----------

(1) Baek M, et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science. 2021
(2) Mariani V, et al. lDDT: a local superposition-free score for comparing protein structures and models using distance difference tests. Bioinformatics. 2013