Returns a hash ignoring the ID - useful for deduplication.

Returns a hash ignoring the ID - useful for deduplication.

Returns a hash ignoring the ID - useful for deduplication.

Return identity Rigid3Array of given shape.

Returns identity of given shape.

Mapping from key values to their index in the column arrays. i.e.: self.key[index_by_key[k]] == k

Run through subbatches (like batch apply but with split and concat).

Encodes a number as a string, using reverse spreadsheet style naming. Args: num: A positive integer. Returns: A string that encodes the

Return Rigid3Array corresponding to inverse transform.

Boolean label indicating if each atom is from DNA or not.

Boolean label indicating if each atom is a ligand or not.

Boolean label indicating if each atom is a nucleic acid or not.

Boolean label indicating if each atom is from protein or not.

Boolean label indicating if each atom is from RNA or not.

Returns if chain type is a protein, DNA or RNA chain type. Args: chain_type: The type of the chain. Returns: A bool for if the chain_ty

Whether hit can be used as a template.

Boolean label indicating if each atom is from water or not.

Iterates pairs of (residue_start, residue_end) indices. Yields: Pairs of (start, end) indices for each residue, where end is not incl

Checks if Jackhmmer supports the --seq-limit flag.

jvp for largest eigenvector.

Lazily parses a FASTA/A3M string and yields (sequence, description) tuples. This implementation is more memory friendly than `fasta_sequence` while

Ratio of the length of the hit sequence to the query.

Returns single letter name if one exists otherwise returns default.

Safely loads pickle data from an already-opened binary file handle. Only built-in container/primitive types listed in `_ALLOWED_BUILTINS` are per

Loads multiple fold inputs from all JSON files in a given input_dir. Args: input_dir: The directory containing the JSON files. Yields: T

Returns a new structure consisting of empty array fields.

Make a flat atom layout for given residues. Create a flat layout from a `Residues` object. The required atoms for each amino acid type are taken

Returns a Structure from a flat layout and atom coordinates. The provided flat_layout must be 1-dim and must not contain any padding elements. Th

Masked mean.

Returns the matching hit sequence including insertions. Make deleted residues uppercase and remove gaps ("-").

Merges chains in each group into a single chain. If a Structure has chains A, B, C, D, E, and `merge_chains([[A, C], [B, D], [E]])` is called

Merges MSA features with shape (NUM_SEQ, NUM_RES) across chains.

Concatenates the paired (all_seq) MSA features with the unpaired ones.

Loads model parameters from the model directory.

Creates an rdkit Mol object from a CCD mmcif data block. The atoms are renumbered so that their names are in alphabetical order and these names a

Creates a CCD-like mmcif data block from an rdkit Mol object. Only a subset of associated mmcif fields is populated, but that is sufficient for f

Compute a residue index array that is monotonic across all chains. Lots of metrics (lddt, l1_long, etc) require computing a distance-along-chain

Returns a config for filter that keeps everything.

Returns Fourier noise level embeddings for diffusion model.

The number of models of this Structure.

The number of models of this Structure.

Returns the final template sequence.

PAE aggregate metrics.

Parses a CIF string with multiple data records. For instance, the CIF string: ``` data_001 _foo bar # data_002 _foo baz ``` is pa

Determines which transforms to apply based on an MMCIF oper_expression str. Args: oper_expression: the field oper_expression from MMCIF format

Compute 1D PDE summaries. Args: num_tokens: The number of tokens (not including padding). asym_ids: The asym_ids (array of shape num_tokens

Populates missing data for the chemical components from CCD. Args: ccd: The chemical components database. chemical_components_data: Chemica

Validate instance has same shapes & dtypes.

Returns table of chains which have at least 1 resolved atom.

Returns table of residues which have at least 1 resolved atom.

Create pseudo beta atom positions and optionally mask. Args: aatype: [num_res] amino acid types. dense_atom_positions: [num_res, NUM_DENSE,

Runs the MSA tool on the target sequence.

Query the database (sharded or unsharded) using Jackhmmer.

0-based query index to hit index mapping.

Query the database using hmmsearch using a given a3m.

Samples uniform random Rot3Array according to Haar Measure.

Compute the metric that will be used to rank predictions, higher is better. Args: full_pde: A [num_samples, num_tokens,num_tokens] matrix of pr

Returns an RDKit Mol, created via RDKit from entry SMILES string.

Realigns the hit sequence to the Structure sequence. For example, for the given input: query_sequence : ABCDEFGHIJKL hit_sequence : ---DE