Utils

Helpful utility functions for the multidms package

multidms.utils.explode_params_dict(params_dict)

Given a dictionary of model parameters, of which any of the values can be a list of values, compute all combinations of model parameter sets and returns a list of dictionaries representing each of the parameter sets.

multidms.utils.my_concat(dfs_list, axis=0)

simple pd.concat wrapper for bug with vscode pylance See https://github.com/matsengrp/multidms/issues/156 for more details

multidms.utils.split_sub(sub_string)

A very simplistic function to split a mutations string into its constituent parts. This function is only designed to work with simple mutation strings of the form A123B, where A is the wildtype amino acid, 123 is the site, and B is the mutant amino acid.

Note

It is favorable to use the polyclonal.utils.MutationParser class instead, as that provides the ability to parse more complex mutation strings using allowed alphabets.

Parameters:

sub_string (str) – A string containing a single mutation

Returns:

A tuple containing the wildtype amino acid, site, and mutant amino acid

Return type:

tuple

multidms.utils.split_subs(subs_string, parser=<function split_sub>)

Given a mutation parsing function, split a string of mutations into three lists containing the wildtype amino acids, sites, and mutant amino acids.

Parameters:

• subs_string (str) – A string containing multiple mutations

• parser (function) – A function that can parse a single mutation string and returns a tuple containing the wildtype amino acid, site, and mutant amino acid

Returns:

A tuple containing the wildtype amino acids, sites, and mutant amino acids as lists

Return type:

tuple

multidms.utils.difference_matrix(n, ref_index=0)

Given some number of conditions, return the difference matrix for computing shifts between adjacent conditional beta parameters.

Parameters:

• n (int) – The number of conditions

• ref_index (int) – The index of the reference condition

Returns:

A difference matrix for computing shifts between adjacent conditional beta parameters

Return type:

jnp.ndarray

multidms.utils.transform(params, bundle_idxs)

Transforms the beta coefficient parameters of a multidms model to be negative for the bundles specified in bundle_idxs, and the updated beta0 parameters based on the new beta coefficients. See issue #156 for more on scaling parameters for training.

Parameters:

• params (dict) – A dictionary containing the model parameters “beta”, and “beta0”.

• bundle_idxs (dict) – A dictionary, keyed by condition containing the bundle indices in the binarymap matrix.

Returns:

A dictionary containing the transformed model parameters “beta”, and “beta0”.

Return type:

dict

multidms.utils.rereference(X, bundle_idxs)

Given a binary matrix X and bundle indices, re-reference the matrix to flip the bit signs for the bundles specified in bundle_idxs. This function is used to scale the data matrix for training a model. See issue #156 for more on scaling parameters for training.

Parameters:

• X (jax.experimental.sparse.BCOO) – A binary matrix

• bundle_idxs (jnp.ndarray) – An boolean array indicating the bundle indices in the binarymap matrix.

Returns:

A re-referenced binary matrix

Return type:

jax.experimental.sparse.BCOO