Returns a copy of this structure with residue IDs remapped. Example structure with 2 chains: Chain A: residues 1, 2, 3; chain B: residues 6,

Removes all gapped rows from all_seq feat based on selected asym_ids.

Remove features of types we don't want to send to the TPU e.g. strings.

Returns a new structure with renamed author chain IDs (auth_asym_ids). Args: new_id_by_old_id: A mapping from original auth_asym_ids to the

Returns a new structure with renamed chains to match another's. Example: This structure has chains: {'A': 'DEEP', 'B': 'MIND', 'C': 'MIND'}

Returns a copy of this structure with residues renamed. Residue names in chemical components data will also be renamed. Args: res_name

Returns the residue auth seq ID for each atom in the structure.

The indices in the atom fields where each present residue begins.

Returns the residue insertion code for each atom in the structure.

Remove author chain/residue ids, entity info and use internal ids.

Create a Residues object from a Structure object.

Residues table.

Computes the *aligned* RMSD of two Mx3 np arrays of coordinates. Args: decoy_coords: [M, 3] np array of decoy atom coordinates. gt_coords:

Launches a subprocess, times it, and checks for errors. Args: cmd: Command to launch. cmd_name: Human-readable command name to be used in l

Sample using denoiser on batch. Args: denoising_step: the denoising function. batch: the batch key: random key config: config for t

Scale translation in Rigid3Array by 'factor'.

Returns a single-letter sequence, taking modifications into account. Uses 'X' for all unknown residues.

Returns a single-letter sequence, taking modifications into account. Uses 'N' for all unknown residues.

Returns a single-letter sequence, taking modifications into account. Uses 'N' for all unknown residues.

Sharded vmap. Maps `fun` over axes, in a way similar to hk.vmap, but does so in shards of `shard_size`. This allows a smooth trade-off between me

Shuffle MSA randomly, return batch with shuffled MSA. Args: key: rng key for random number generation. msa: MSA object to sample msa from.

The number of rows in the table.

Used to create a new Structure by slicing into the leading dimensions. Example usage 1: ``` final_state = multi_state_struc.slice_leadin

Splits a Structure into single-chain Structures, one for each chain. The obtained structures can be merged back together into the original st

Stacks multiple structures into a single multi-model Structure. This function is the inverse of `Structure.unstack()`. NB: this function assumes

Encodes an mmCIF-style string chain ID as an integer. The integer IDs are one based so this function is the inverse of int_id_to_str_id. Args:

Yields template structures for each unique PDB ID among hits. If there are multiple hits in the same Structure, the Structure will be include

The names of the tables in this database.

Returns Structure tables constructed from atom array level data. All fields except name and, res_id are optional, all array fields consist of a v

Returns all targets in the store.

Run featurisation and assert that the output is as expected.

Run AlphaFold 3 inference.

Run model inference and assert that output exists.

Test that the db_dir is replaced correctly.

Convert Rot3Array to array of shape [..., 3, 3].

Converts ProteinChain to an AlphaFold JSON dict.

Converts RnaChain to an AlphaFold JSON dict.

Converts DnaChain to an AlphaFold JSON dict.

Converts StructureConfidenceFull to json string.

Returns raw _atom_site and _struct_conn mmCIF tables.

Returns chemical components data as a dict suitable for `mmcif.Mmcif`.

Returns the bioassembly data as a dict suitable for `mmcif.Mmcif`.

Returns raw sequence and entity mmCIF tables.

Returns list of atom dictionaries grouped by residue. If this is a multi-model structure, each atom will store its fields atom_x, atom_y, ato

Maps a flat atom layout to tokens for evoformer. Creates the evoformer tokens as one token per polymer residue and one token per ligand atom. The

Transforms states to chains. A multi-state protein structure will be transformed to a multi-chain single-state protein structure. Useful for

Returns the element type for the given component name and atom name. Args: ccd: The chemical components dictionary. res_name: The component

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

Walks the directory tree of resources similar to os.walk.

Computes weighted mean in a safe way that avoids NaNs. This is equivalent to jnp.average for the case eps=0.0, but adds a small constant to the d

Returns a new structure with the chains are in reverse spreadsheet style. This is the usual order to write chains in an mmCIF: (A < B < ... <

Writes embeddings to a directory.

Writes processed inference result to a directory.