calculate dihedral angles for all consecutive quadruples (a[i],b[i],c[i],d[i]) given Cartesian coordinates of four sets of atoms a,b,c,d Para

Helper function to go from discrete time index t to corresponding sigma_idx. Args: t: time index (integer between 1 and 2

Given a noise schedule type, create the beta schedule

Multiply two quaternions. Usual torch rules for broadcasting apply. Args: a: Quaternions as tensor of shape (..., 4), real part

Extract \sigma(t) corresponding to chosen sigma schedule. Args: t: torch tensor with time between 0 and 1

Extract the first or third Euler angle from the two members of the matrix which are positive constant times its sine and cosine. Args:

Cosine interpolation of some value between its max <eta_max> and its min <eta_min> from https://pytorch.org/docs/stable/generated/torch.opti

function to generate inpaint_str or inpaint_seq masks specific to this contig

extract xyz coords for all heavy atoms

Convert rotations given as quaternions to rotation matrices. Args: quaternions: quaternions with real part first, as ten

sample uses the inverse cdf to sample an angle of rotation from IGSO(3) Args: ts: array of integer time steps to sample fr

Return the rotation matrices for one of the rotations about an axis of which Euler angles describe, for each value of the angle given. A

_calc_igso3_vals computes numerical approximations to the relevant analytically intractable functionals of the igso3 distribution.

Return a tensor where each element has the absolute value taken from the, corresponding element of a, with sign taken from the corresponding

Returns torch.sqrt(torch.max(0, x)) but with a zero subgradient where x is 0.

Need to align px0 to motif in xT. This is to permit the swapping of residue positions in the px0 motif for the true coordinates. Firs

Applies a noising kernel to the points in x Parameters: x (torch.tensor, required): (N,3,3) set of backbone coordinates

Repeatedly apply self.apply_kernel T times and return all crds

Convert rotations given as axis/angle to quaternions. Args: axis_angle: Rotations given as a vector in axis angle form,

Calculates number of chains for given symmetry

score uses the quotient rule to compute the scaling factor for the score of the IGSO(3) density. This function is used within the Diffuser cl

Given the current sequence and structure of the model prediction, return the current potential as a PyTorch tensor with a sin

This is the money call. Take the current sequence and structure information and get the sum of all of the potentials that are being used

Create contig from target features.

Make length-specific denoiser.

IGSO(3) density. Args: expansion: truncated approximation of the power series in the IGSO(3) density. omega: length of an

Perform spherical linear interpolation from the True coordinate frame for each residue to a randomly sampled coordinate frame

Given full atom xyz, sequence and atom mask, diffuse the protein translations, rotations, and chi angles Parameters:

exp_score_norm returns the expected value of norm of the score for IGSO(3) with time parameter ts of shape [T].

Function to generate a new mask with dilated loops and N and C terminal additions

Given an expanded mask, populate a new ss and adj based on this

g returns the drift coefficient at time t since sigma(t)^2 := \int_0^t g(s)^2 ds, for arbitrary sigma(t) we invert this

recreate Cb given N,Ca,C

calculate planar angles for all consecutive triples (a[i],b[i],c[i]) from Cartesian coordinates of three sets of atoms a,b,c Parameters

Function to precalculate beta_T for chi angles (decoded at different time steps, so T in beta_T varies). Calculated empirically

Returns the next (0-indexed) positions to decode for this timestep

Given a timestep and a decay type, get the appropriate scale factor to use for applying guiding potentials Inputs:

Given xt, predicted x0 and the timestep t, give mu of x(t-1) Assumes t is 0 indexed

Given full atom x0 prediction (xyz coordinates), diffuse to x(t-1) Parameters: xt (L, 14/27, 3) set of coordinates

get_next_frames gets updated frames using either SLERP or the IGSO(3) + score_based reverse diffusion. based on self.so3_type use slerp or s

Given indices being currently sampled and a distance map, return one more index which is allowed to be sampled at the same time as cu

Wrapper function to take px0, xt and t, and to produce xt-1 First, aligns px0 to xt Then gets coordinates, frames and torsion

calculate pair distances between two sets of points Parameters ---------- a,b : pytorch tensors of shape [batch,nres,3] sto

This could be moved into potential manager if desired - NRB Function to take a structure (x) and get per-atom gradients used to guid

Function to get a sampled mask from a contig.

Given at item, get the ss tensor and block adjacency matrix for that item

Truncated sum of IGSO(3) distribution. This function approximates the power series in equation 5 of "DENOISING DIFFUSION PROBABILISTIC MODELS

Truncated sum of IGSO(3) distribution implemented in torch Args: omega: rotation of Euler vector (i.e. the angle of rotation) sig

Given a list of potential dictionaries where each dictionary defines the configurations for a single potential, initialize al

Check whether this instance of PotentialManager actually contains any potentials

Loads RF checkpoint, from which config can be generated.

Create RosettaFold model from preloaded checkpoint.

Calculate a matrix of inter/intra chain contact indicators Parameters: nchain (int, required): How many chains are in this desig

Given an expanded ss and adj, mask some number of residues at either end of non-loop ss

Takes a mask of True (loop) and False (non-loop), and outputs list of tuples (loop or not, length of element)

Convert rotations given as rotation matrices to quaternions. Args: matrix: Rotation matrices as tensor of shape (..., 3, 3). Re

Parse a single entry in the list of potentials to be run to a dictionary of settings for that potential. An example of how t

Function to prepare inputs to diffusion model. Prepared in a form recognizable by the structure prediction network to enable