calculate dynamic programming matrix using gotoh algorithm * S - cumulative scorefor each cell * P - string representation for path *

Computes aligned confidence metrics from logits. Args: logits: [*, num_res, num_res, num_bins] the logits output from PredictedAlig

Compute several metrics to assess the structural violations.

Entropy loss of unnormalized logits. logits: Affinities are over the last dimension https://github.com/google-research/magvit/blob/05e8cfd65

extract pairwise sequence alignment from residue index vectors, * assuming that "sequence" contains two empty strings. * return length of alignmen

extract pairwise sequence alignment from residue index vectors, * assuming that "sequence" contains two empty strings. * return length of alignmen

Args: structure: An instance of biotite AtomArray atoms: default ["N", "CA", "C"] Returns: Tuple (coords, seq)

Counts residues whose Ca is a large distance from its neighbour. Measures the fraction of CA-CA pairs between consecutive amino acids that are

Filter all peptide backbone atoms of one array. This includes the "N", "CA" and "C" atoms of amino acids. Parameters ---------- arra

locate the cell with highest alignment score. reset path after * the cell to zero */

Runs a forward pass given input tokens. Use `model.infer` to run inference from a sequence. Args: aa (torch.Tensor): Tens

Runs a forward pass given input tokens. Use `model.infer` to run inference from a sequence. Args: aa (torch.Tensor): Tens

Convert frames to their idealized all atom representation. Args: r: All rigid groups. [..., N, 8, 3] aatype: Residue types. [...,

Convert frames to their idealized all atom representation. Args: r: All rigid groups. [..., N, 8, 3] aatype: Residue types. [...,

Example for atoms argument: ["N", "CA", "C"]

perform gapless threading to find the best initial alignment input: x, y, xlen, ylen output: y2x0 stores the best alignment: e.g., y2x0[j]=i means: th

get_initial5 in TMalign fortran, get_initial_local in TMalign c by yangji get initial alignment of local structure superposition input: x, y, xlen, yl

perform fragment gapless threading to find the best initial alignment input: x, y, xlen, ylen output: y2x0 stores the best alignment: e.g., y2x0[j]=i

get initial alignment from secondary structure alignment input: x, y, xlen, ylen output: y2x stores the best alignment: e.g., y2x[j]=i means: the jth

get initial alignment from secondary structure and previous alignments input: x, y, xlen, ylen output: y2x stores the best alignment: e.g., y2x[j]=i m

Initializes multi-GPU-friendly python command line logger.

Runs a forward pass given input sequences. Args: sequences (Union[str, List[str]]): A list of sequences to make predictions for.

Returns list of pdb (files) strings from the model given a list of input sequences.

Returns list of pdb (files) strings from the model given a list of input sequences.

initialize matrix in gotoh algorithm */