pyprobound.experiment.Experiment

class Experiment(rounds, counts_per_round=None, name='')

Bases: Transform

Models sequenced rounds corresponding to count table columns.

observed_rounds

The sequenced rounds modeled by the experiment.

Type:

list[BaseRound]

rounds

A ModuleList of all rounds, including those not sequenced.

Type:

TModuleList[BaseRound]

__init__(rounds, counts_per_round=None, name='')

Initializes the experiment from an iterable of sequenced rounds.

Parameters:

• rounds (Iterable[BaseRound]) – The sequenced rounds modeled by the experiment.

• counts_per_round (Tensor | list[float] | None) – A tensor with the number of probes in each round of the count table used for training, with shape \((\text{rounds},)\). Should be provided for Kd-seq or pseudocount > 0.

• name (str) – A string used to describe the experiment.

Methods

cache(fun)	Decorator for a function to cache its output.
check_length_consistency()	Checks that input lengths of Binding components are consistent.
components()	Iterator of child components.
forward(seqs)	Predicts the log probe frequencies.
free_protein(i_index, b_index, u_index[, ...])	Estimates the free protein concentration.
freeze()	Turns off gradient calculation for all parameters.
max_embedding_size()	The maximum number of bytes needed to encode a sequence.
optim_procedure([ancestry, current_order])	The sequential optimization procedure for all Binding components.
reload(checkpoint)	Loads the model from a checkpoint file.
reload_from_state_dict(state_dict)	Loads the model from a state dict.
save(checkpoint[, flank_lengths])	Saves the model to a file with "state_dict" and "metadata" fields.
unfreeze([parameter])	Turns on gradient calculation for the specified parameter.

Attributes

counts_per_round	A list of the number of probes in each round of the count table used for training, with shape \((\text{rounds},)\).
unfreezable	alias of Literal['all']

Non-Inherited Members

property counts_per_round: list[float]

A list of the number of probes in each round of the count table used for training, with shape \((\text{rounds},)\).

components()

Iterator of child components.

Return type:

Iterator[BaseRound]

forward(seqs)

Predicts the log probe frequencies.

\[\log \frac{\eta_{r} f_{i,r}}{ \sum_{r^\prime} \eta_{r^\prime} f_{i, r^\prime} }\]

Parameters:

seqs (Tensor) – A sequence tensor of shape \((\text{minibatch},\text{length})\) or \((\text{minibatch},\text{in_channels},\text{length})\).

Return type:

Tensor

Returns:

The log frequency tensor of shape \((\text{minibatch},\text{rounds})\).

free_protein(i_index, b_index, u_index, target_concentration=None, library_concentration=None)

Estimates the free protein concentration.

If the input (I), bound (B), and unbound (U) probes of a selection are modeled jointly, PyProBound can estimate free protein concentration.

\[\begin{split}[\text{P}]_F &= [\text{P}]_T - [\text{library}] p(B) \\ p(B) &= \frac{k_B}{k_I} \frac{\eta_I}{\eta_B}\end{split}\]

Parameters:

• i_index (int) – The index of the input round.

• b_index (int) – The index of the BoundRound.

• u_index (int) – The index of the UnboundRound.

• target_concentration (float | None) – The total protein concentration \([\text{P}]_T\), taken from Aggregate if not provided.

• library_concentration (float | None) – The total library concentration \([\text{library}]\), taken from Round if not provided.

Return type:

float

Returns:

The free protein concentration represented as a float.