Tools

This module contains a series of core tools used for the feature calculations in DeepRank. These tools are at the moment:

• sasa: a simple solvent surface area calculator
• sparse: a 3D sparse matrix engine

Here are the details of the submodule given in tools.

Solvent Accessible Surface Area

class deeprank.tools.sasa.SASA(pdbfile)

Simple class that computes Surface Accessible Solvent Area.

The method follows some of the approaches presented in:

Solvent accessible surface area approximations for rapid and accurate protein structure prediction https://link.springer.com/article/10.1007%2Fs00894-009-0454-9

Example

>>> sasa = SASA('1AK4_1w.pdb')
>>> NV = sasa.neighbor_vector()
>>> print(NV)

Parameters:pdbfile (str) – PDB file of the conformation

get_center(chain1='A', chain2='B', center='cb')

Get the center of the resiudes.

Parameters:

• chain1 (str, optional) – Name of the first chain
• chain2 (str, optional) – Name of the second chain
• center (str, optional) – Specify the center. ‘cb’: the center locates on carbon beta of each residue ‘center’: average position of all atoms of the residue

Raises:

ValueError – If the center is not recpgnized

get_residue_center(chain1='A', chain2='B')

Compute the average position of all the residues.

Parameters:

• chain1 (str, optional) – Name of the first chain
• chain2 (str, optional) – Name of the second chain

get_residue_carbon_beta(chain1='A', chain2='B')

Extract the position of the carbon beta of each residue.

Parameters:

• chain1 (str, optional) – Name of the first chain
• chain2 (str, optional) – Name of the second chain

neighbor_vector(lbound=3.3, ubound=11.1, chain1='A', chain2='B', center='cb')

Compute teh SASA folowing the neighbour vector approach.

The method is based on Eq on page 1097 of https://link.springer.com/article/10.1007%2Fs00894-009-0454-9

Parameters:

• lbound (float, optional) – lower boubd
• ubound (float, optional) – upper bound
• chain1 (str, optional) – name of the first chain
• chain2 (str, optional) – name of the second chain
• center (str, optional) – specify the center (see get_residue_center)

Returns:

neighbouring vectors

Return type:

dict

neighbor_count(lbound=4.0, ubound=11.4, chain1='A', chain2='B', center='cb')

Compute the neighbourhood count of each residue.

The method is based on Eq on page 1097 of https://link.springer.com/article/10.1007%2Fs00894-009-0454-9

Parameters:

• lbound (float, optional) – lower boubd
• ubound (float, optional) – upper bound
• chain1 (str, optional) – name of the first chain
• chain2 (str, optional) – name of the second chain
• center (str, optional) – specify the center
• get_residue_center) ((see) –

Returns:

Neighborhood count

Return type:

dict

static neighbor_weight(dist, lbound, ubound)

Neighboor weight.

Parameters:

• dist (np.array) – distance from neighboors
• lbound (float) – lower bound
• ubound (float) – upper bound

Returns:

distance

Return type:

float

Sparse 3D Matrix

deeprank.tools.sparse._printif(string, cond)

class deeprank.tools.sparse.FLANgrid(sparse=None, index=None, value=None, shape=None)

Flat Array sparse matrix.

Parameters:

• sparse (bool, optional) – Sparse or Not
• index (list(int), optional) – single index of each non-zero element
• value (list(float), optional) – values of non-zero elements
• shape (3x3 array, optional) – Shape of the matrix

from_dense(data, beta=None, debug=False)

Create a sparse matrix from a dense one.

Parameters:

• data (np.array) – Dense matrix
• beta (float, optional) – threshold to determine if a sparse rep is valuable
• debug (bool, optional) – print debug information

to_dense(shape=None)

Create a dense matrix.

Parameters:shape (3x3 array, optional) – Shape of the matrix Returns:Dense 3D matrix Return type:np.array Raises:ValueError – shape not defined

_get_single_index(index)

Get the single index for a single element.

# get the index can be used with a map # self.index = np.array( list( map(self._get_single_index,index) ) ).astype(index_type) # however that is remarkably slow compared to the array version

Parameters:index (array) – COO index Returns:index Return type:int

_get_single_index_array(index)

Get the single index for multiple elements.

# get the index can be used with a map # self.index = np.array( list( map(self._get_single_index,index) ) ).astype(index_type) # however that is remarkably slow compared to the array version

Parameters:index (array) – COO index Returns:index Return type:list(int)