utils module

utils.atomlistToChannels(atomNames, hashing='Element_Hashing', device='cpu')[source]

function to get channels from atom names (obtained parsing the pdb files with the parsePDB function)

Parameters

  • atomNames (list) – atom names obtained parsing the pdb files with the parsePDB function

  • hashing ("TPL_Hashing" or "Element_Hashing" or dict) –

    define which atoms are grouped together. You can use two default hashings or build your own hashing:

    TPL_Hashing: uses the hashing of torch protein library (https://github.com/lupoglaz/TorchProteinLibrary) Element_Hashing: groups atoms in accordnce with the element only: C -> 0, N -> 1, O ->2, P ->3, S- >4, H ->5, everything else ->6

    Alternatively, if you are not happy with the default hashings, you can build a dictionary of dictionaries that defines the channel of every atom type in the pdb. the first dictionary has the residue tag (three letters amino acid code) as key (3 letters compound name for hetero atoms, as written in the PDB file) every residue key is associated to a dictionary, which the atom tags (as written in the PDB files) as keys and the channel (int) as value

    for example, you can define the channels just based on the atom element as following: { ‘CYS’: {‘N’: 1, ‘O’: 2, ‘C’: 0, ‘SG’: 3, ‘CB’: 0, ‘CA’: 0}, # channels for cysteine atoms ‘GLY’: {‘N’: 1, ‘O’: 2, ‘C’: 0, ‘CA’: 0}, # channels for glycine atom … ‘GOL’: {‘O1’:2,’O2’:2,’O3’:2,’C1’:0,’C2’:0,’C3’:0}, # channels for glycerol atom … }

    The default encoding is the one that assigns a different channel to each element

    other encodings can be found in sources/hashings.py

  • device (torch.device) – The device on which the model should run. E.g. torch.device(“cuda”) or torch.device(“cpu:0”)

Returns

  • coords (torch.Tensor) – coordinates of the atoms in the pdb file(s). Shape ( batch, numberOfAtoms, 3)

  • channels (torch.tensor) – the channel of every atom. Shape (batch,numberOfAtoms)

utils.atomlistToRadius(atomList, hashing='FoldX_radius', device='cpu')[source]

function to get radius from atom names (obtained parsing the pdb files with the parsePDB function)

Parameters

  • atomNames (list) – atom names obtained parsing the pdb files with the parsePDB function

  • hashing (FoldX_radius or dict) –

    “FoldX_radius” provides the radius used by the FoldX force field

    Alternatively, if you are not happy with the foldX radius, you can build a dictionary of dictionaries that defines the radius of every atom type in the pdb. The first dictionary has the residue tag (three letters amino acid code) as key (3 letters compound name for hetero atoms, as written in the PDB file) every residue key is associated to a dictionary, which the atom tags (as written in the PDB files) as keys and the radius (float) as value

    for example, you can define the radius as following: { ‘CYS’: {‘N’: 1.45, ‘O’: 1.37, ‘C’: 1.7, ‘SG’: 1.7, ‘CB’: 1.7, ‘CA’: 1.7}, # radius for cysteine atoms ‘GLY’: {‘N’: 1.45, ‘O’: 1.37, ‘C’: 1.7, ‘CA’: 1.7}, # radius for glycine atoms … ‘GOL’: {‘O1’:1.37,’O2’:1.37,’O3’:1.37,’C1’:1.7,’C2’:1.7,’C3’:1.7}, # radius for glycerol atoms … }

    The default radius are the ones defined in FoldX

    Radius default dictionary can be found in sources/hashings.py

  • device (torch.device) – The device on which the model should run. E.g. torch.device(“cuda”) or torch.device(“cpu:0”)

Returns

  • coords (torch.Tensor) – coordinates of the atoms in the pdb file(s). Shape ( batch, numberOfAtoms, 3)

  • radius (torch.tensor) – The radius of every atom. Shape (batch,numberOfAtoms)

utils.parsePDB(PDBFile, keep_only_chains=None, keep_hetatm=True, bb_only=False)[source]

function to parse pdb files. It can be used to parse a single file or all the pdb files in a folder. In case a folder is given, the coordinates are gonna be padded

Parameters

  • PDBFile (str) – path of the PDB file or of the folder containing multiple PDB files

  • bb_only (bool) – if True ignores all the atoms but backbone N, C and CA

  • keep_only_chains (str or None) – ignores all the chain but the one given. If None it keeps all chains

  • keep_hetatm (bool) – if False it ignores heteroatoms

Returns

  • coords (torch.Tensor) – coordinates of the atoms in the pdb file(s). Shape ( batch, numberOfAtoms, 3)

  • atomNames (list) – a list of the atom identifier. It encodes atom type, residue type, residue position and chain

utils.parseSDF(SDFFile)[source]

function to parse pdb files. It can be used to parse a single file or all the pdb files in a folder. In case a folder is given, the coordinates are gonna be padded

Parameters

SDFFile (str) – path of the PDB file or of the folder containing multiple PDB files

Returns

  • coords (torch.Tensor) – coordinates of the atoms in the pdb file(s). Shape ( batch, numberOfAtoms, 3)

  • atomNames (list) – a list of the atom identifier. It encodes atom type, residue type, residue position and chain