pyrosetta_help.alphafold package

Submodules

pyrosetta_help.alphafold.constraints module

pyrosetta_help.alphafold.constraints.add_interchain_pae_constraints(pose, errors, cutoff=15)

Add constraints between residues that are interacting according to the PAE error matrix but are in different ‘chains’ (sensu PyRosetta FoldTree).

Parameters:

• pose –

• errors –

• cutoff –

Returns:

pyrosetta_help.alphafold.constraints.add_pae_constraints(pose: Pose, errors: ndarray, cutoff: float = 12, tolerance: float | None = None, weight: float = 1, adjecency_threshold=5) → None

Add constrains to the pose based on the errors matrix. NB. this matrix is a reshaped version of what AF2 returns.

A harmonic function is added to CA atoms that are in residues with the error under a specified cutoff. The mu is the current distance and the standard deviation of the harmonic is the error times weight.

To find out how many were added:

>>> len(pose.constraint_set().get_all_constraints())

Parameters:

• pose –

• errors –

• cutoff –

• tolerance – if None Harmonic, if value, tollerance of FlatHarmonic

• weight – this is added to the SD part so squared inverse.

• adjecency_threshold – min residue separation of sequence neighbours

Returns:

pyrosetta_help.alphafold.constraints.add_stretch_constraint(pose: Pose, weight: float = 5, slope_in: float = -0.05, residue_index_A: int = 1, residue_index_B: int = -1, distance: float | None = None, sigmoid: bool = True) → AtomPairConstraint

Add a constraint to “stretch out” the model, because slope_in is negative. The weight needs to be negative for sigmoid=False or it will attractive

Parameters:

• pose – Pose to add constraint to

• weight – how strength of constraint (max of 0.5 for SigmoidFunc)

• slope_in – negative number to stretch

• residue_index_A – first residue?

• residue_index_B – last residue is “-1”

• distance – if omitted, the midpoint of Sigmoid will be the current distance

• sigmoid – use sigmoid or identity/linear (bad idea)

Returns:

pyrosetta_help.alphafold.constraints.make_pae_constraint(pose, residue1_pose_idx: int, residue2_pose_idx: int, error: float, tolerance: float | None = None, weight: float = 1)

Add a constraint between two residues based on the PAE error from AlphaFold2 (the colourful heatmap in EBI-AF2).

Parameters:

• pose –

• residue1_pose_idx –

• residue2_pose_idx –

• error –

• tolerance –

• weight –

Returns:

pyrosetta_help.alphafold.multimodel module

class pyrosetta_help.alphafold.multimodel.AF2NotebookAnalyser(folder: str, load_poses: bool = True)

Bases: object

This class features heavily in the notebook colab-pyrosetta-dimer.ipynb It allows the analysis of the results of the AF2 protocol for a dimer. Namely, .sidechain_relax

analyser = ph.AF2NotebookAnalyser(folder=folder_name, load_poses=True)
analyser.sidechain_relax(cycles)
analyser.constrain(tolerance=2)
analyser.relax(cycles)
analyser.calculate_interface()
score: pd.DataFrame = analyser.scores

The .scores attribute is a pandas DataFrame generated by .make_AF2_dataframe

One can also make a phosphorylated model

pdb_ptms = analyser.parse_phosphosite(raw, maximum=analyser.original_poses[1].chain_end(1))
analyser.make_phosphorylated(pdb_ptms, chain, cycles)

To access specific poses the following attributes are available:

print(analyser.pose_groupnames) # ['relaxed', 'original', 'phospho']
analyser.original_poses[i]
analyser.relaxed_poses[i]
analyser.phospho_poses[i]

__init__(folder: str, load_poses: bool = True)

calculate_interface(interface='A_B')

Calculates the interface strength between the two chains.

It does a median of the pLDDT scores of the interface residues.

Parameters:

interface – A Rosetta format string for the description of the interface, e.g. ‘A_B’

Returns:

constrain(groupname: str = 'relaxed', **add_pae_constraints_arguments)

Calls the add_pae_constraints and add_interchain_pae_constraints functions on each pose in the group. :param groupname: :param add_pae_constraints_arguments: :return:

constrain_and_relax(cycles: int = 3)

Adds constraints and relaxes the poses.

dump(folder: None | str = None)

Saves ALL the pdbs (via dump_pdbs) and the errors.

dump_pdbs(groupname: str = 'relaxed', folder: None | str = None, prefix: str | None = '')

Save in the format f'{prefix}rank_{index}_pyrosetta_{groupname}.pdb'.

find_interface_residues()

Calls get_interactions for each pose and saves the result in the scores dataframe.

get_errors() → dict

The errors are the result of reshaping via ph.reshape_errors :return:

get_interactions(pose: Pose, chain_id: int, threshold: float = 3.0) → ResidueVector

Get the sequence of pose residue indices that are within the threshold distance of the chain_id.

get_median_interface_bfactors() → Dict[int, float]

“bfactors” aren’t actually bfactors, but pLDDT values. Based on what are the residues identified with find_interface_residues in calculates the median pLDDT of these.

get_poses() → Dict[int, Pose]

This used to be a pyrosetta.rosetta.utility.vector1_core_pose_Pose but it turns out that the getitem of vector1_core_pose_Pose returns a clone of the pose, not the pose itself.

Returns:

classmethod load(folder: str, load_poses=False, params=())

Loads the poses (from PDB files), scores and errors from the folder. :return:

make_AF2_dataframe() → DataFrame

Given a folder form ColabsFold return a dictionary with key rank index and value a dictionary of details

This is convoluted, but it may have been altered by a human.

make_phosphorylated(pdb_ptms: dict, chain: str = 'A', cycles: int = 3)

Make a phosphorylated pose from the dictionary of Phoshositeplus annotations (see parse_phosphosite)

classmethod parse_phosphosite(raw: str, minimum: int = 1, maximum: int = -1)

A rubbish method to convert copy-pasted Phosphosite web table.

Parameters:

• raw –

• minimum –

• maximum –

Returns:

property poses

relax(cycles: int = 3)

This does backbone and sidechain relaxation. For only sidechain relaxation, use .sidechain_relax.

Parameters:

cycles –

Returns:

sidechain_relax(cycles: int = 5)

This is a prelude to full relaxation. This is relaxes only the sidechains, which if strained or clashing may result in the whole pose blowing up.

Parameters:

cycles – cycles of FastRelax

Returns:

pyrosetta_help.alphafold.plot module

pyrosetta_help.alphafold.plot.make_pae_plot(errors: ndarray) → plotly.import graph_objs.Figure

Make AlphaFold2-EBI–like PAE plot (green and white)

pyrosetta_help.alphafold.retrieval module

pyrosetta_help.alphafold.retrieval.get_alphafold2_error(uniprot: str, reshaped=True) → ndarray | list

Returns the distances errors either as numpy matrix (reshaped=True) or as the weird format from AF2-EBI —see help(pyrosetta_help.alphafold.retrieval.reshape_errors) for more.

Remember that the matrix is zero indexed and that these values are in Ångström and are not pLDDT, which are stored as b-factors.

pyrosetta_help.alphafold.retrieval.pose_from_alphafold2(uniprot: str) → Pose

Returns a pose from the alphafold2 server.

Parameters:

uniprot – uniprot id (“accession”), not gene name or uniprot name.

Returns:

pyrosetta_help.alphafold.retrieval.reshape_errors(errors: List[Dict[str, list]]) → array

The JSON from AF2 has a single element list. the sole element is a dictionary with keys ‘residue1’, ‘residue2’ and ‘distance’. This method returns a matrix of distances reshaped based on the stated residue indices. This is rather unlikely to differ from a regular reshape… but idiotically I am not taking changes assuming it is always sorted.

pyrosetta_help.alphafold.superimpose module

pyrosetta_help.alphafold.superimpose.superimpose_by_pLDDT(pose: Pose, original: Pose, cutoff=70, pose_range=None) → map_core_id_AtomID_core_id_AtomID

Superimpose two poses, based on residues with pLDDT above a given threshold.

Parameters:

• pose –

• original –

• cutoff –

  %

• pose_range – optional argument to subset (start:int, end:int)

Returns:

Module contents