openfold.data.data_transforms¶

Functions

`add_constant_field`(protein, key, value)
`add_distillation_flag`(protein, distillation)
`atom37_to_frames`(protein[, eps])
`atom37_to_torsion_angles`(protein[, prefix])	Convert coordinates to torsion angles.
`block_delete_msa`(protein, config)
`cast_to_64bit_ints`(protein)
`correct_msa_restypes`(protein)	Correct MSA restype to have the same order as rc.
`crop_extra_msa`(protein, max_extra_msa)
`crop_templates`(protein, max_templates)
`curry1`(f)	Supply all arguments but the first.
`delete_extra_msa`(protein)
`fix_templates_aatype`(protein)
`get_backbone_frames`(protein)
`get_chi_angles`(protein)
`get_chi_atom_indices`()	Returns atom indices needed to compute chi angles for all residue types.
`make_all_atom_aatype`(protein)
`make_atom14_masks`(protein)	Construct denser atom positions (14 dimensions instead of 37).
`make_atom14_masks_np`(batch)
`make_atom14_positions`(protein)	Constructs denser atom positions (14 dimensions instead of 37).
`make_fixed_size`(protein, shape_schema, ...)	Guess at the MSA and sequence dimension to make fixed size.
`make_hhblits_profile`(protein)	Compute the HHblits MSA profile if not already present.
`make_masked_msa`(protein, config, ...)	Create data for BERT on raw MSA.
`make_msa_feat`(protein)	Create and concatenate MSA features.
`make_msa_mask`(protein)	Mask features are all ones, but will later be zero-padded.
`make_one_hot`(x, num_classes)
`make_pseudo_beta`(protein[, prefix])	Create pseudo-beta (alpha for glycine) position and mask.
`make_seq_mask`(protein)
`make_template_mask`(protein)
`nearest_neighbor_clusters`(protein[, ...])
`pseudo_beta_fn`(aatype, all_atom_positions, ...)	Create pseudo beta features.
`random_crop_to_size`(protein, crop_size, ...)	Crop randomly to crop_size, or keep as is if shorter than that.
`randomly_replace_msa_with_unknown`(protein, ...)	Replace a portion of the MSA with 'X'.
`sample_msa`(protein, max_seq, keep_extra[, seed])	Sample MSA randomly, remaining sequences are stored are stored as extra_*.
`sample_msa_distillation`(protein, max_seq)
`select_feat`(protein, feature_list)
`shaped_categorical`(probs[, epsilon])
`squeeze_features`(protein)	Remove singleton and repeated dimensions in protein features.
`summarize_clusters`(protein)	Produce profile and deletion_matrix_mean within each cluster.
`unsorted_segment_sum`(data, segment_ids, ...)	Computes the sum along segments of a tensor.

add_constant_field(protein, key, value)¶

add_distillation_flag(protein, distillation)¶

atom37_to_frames(protein, eps=1e-08)¶

atom37_to_torsion_angles(protein, prefix='')¶

Convert coordinates to torsion angles.

This function is extremely sensitive to floating point imprecisions and should be run with double precision whenever possible.

Parameters:

containing (Dict) –

(prefix)aatype:
[*, N_res] residue indices
(prefix)all_atom_positions:
[*, N_res, 37, 3] atom positions (in atom37 format)
(prefix)all_atom_mask:
[*, N_res, 37] atom position mask

Returns:

“(prefix)torsion_angles_sin_cos” ([*, N_res, 7, 2]): Torsion angles
”(prefix)alt_torsion_angles_sin_cos” ([*, N_res, 7, 2]): Alternate torsion angles (accounting for 180-degree symmetry)
”(prefix)torsion_angles_mask” ([*, N_res, 7]): Torsion angles mask

Return type:

The same dictionary updated with the following features

block_delete_msa(protein, config)¶

cast_to_64bit_ints(protein)¶

correct_msa_restypes(protein)¶: Correct MSA restype to have the same order as rc.

crop_extra_msa(protein, max_extra_msa)¶

crop_templates(protein, max_templates)¶

curry1(f)¶: Supply all arguments but the first.

delete_extra_msa(protein)¶

fix_templates_aatype(protein)¶

get_backbone_frames(protein)¶

get_chi_angles(protein)¶

get_chi_atom_indices()¶

Returns atom indices needed to compute chi angles for all residue types.

Returns:: A tensor of shape [residue_types=21, chis=4, atoms=4]. The residue types are in the order specified in rc.restypes + unknown residue type at the end. For chi angles which are not defined on the residue, the positions indices are by default set to 0.

make_all_atom_aatype(protein)¶

make_atom14_masks(protein)¶: Construct denser atom positions (14 dimensions instead of 37).

make_atom14_masks_np(batch)¶

make_atom14_positions(protein)¶: Constructs denser atom positions (14 dimensions instead of 37).

make_fixed_size(protein, shape_schema, msa_cluster_size, extra_msa_size, num_res=0, num_templates=0)¶: Guess at the MSA and sequence dimension to make fixed size.

make_hhblits_profile(protein)¶: Compute the HHblits MSA profile if not already present.

make_masked_msa(protein, config, replace_fraction, seed)¶: Create data for BERT on raw MSA.

make_msa_feat(protein)¶: Create and concatenate MSA features.

make_msa_mask(protein)¶: Mask features are all ones, but will later be zero-padded.

make_one_hot(x, num_classes)¶

make_pseudo_beta(protein, prefix='')¶: Create pseudo-beta (alpha for glycine) position and mask.

make_seq_mask(protein)¶

make_template_mask(protein)¶

nearest_neighbor_clusters(protein, gap_agreement_weight=0.0)¶

pseudo_beta_fn(aatype, all_atom_positions, all_atom_mask)¶: Create pseudo beta features.

random_crop_to_size(protein, crop_size, max_templates, shape_schema, subsample_templates=False, seed=None)¶: Crop randomly to crop_size, or keep as is if shorter than that.

randomly_replace_msa_with_unknown(protein, replace_proportion)¶: Replace a portion of the MSA with ‘X’.

sample_msa(protein, max_seq, keep_extra, seed=None)¶: Sample MSA randomly, remaining sequences are stored are stored as extra_*.

sample_msa_distillation(protein, max_seq)¶

select_feat(protein, feature_list)¶

shaped_categorical(probs, epsilon=1e-10)¶

squeeze_features(protein)¶: Remove singleton and repeated dimensions in protein features.

summarize_clusters(protein)¶: Produce profile and deletion_matrix_mean within each cluster.

unsorted_segment_sum(data, segment_ids, num_segments)¶

Computes the sum along segments of a tensor. Similar to tf.unsorted_segment_sum, but only supports 1-D indices.

Parameters:

data – A tensor whose segments are to be summed.
segment_ids – The 1-D segment indices tensor.
num_segments – The number of segments.

Returns:

A tensor of same data type as the data argument.