API¶

Data¶

Basic routines

`list_contexts`()	List registered contexts.
`describe_context`(context_id)	Describe specified context.
`register_context`(context_class)	Register new context to the collection.
`load_anndata`(context_id[, hgnc_renaming])	Load data from given context as `AnnData`.
`load_plibdata`(contexts_ids[, ...])	Load data from given context(s) as `PlibData`.
`split_plibdata_2fold`(pdata, context_ids)	Split data to training and validation.
`split_plibdata_3fold`(pdata, context_ids)	Split data to training, validation, and test.

Data structures

`PlibData`	Data structure for hosting perturb-lib data.
`InMemoryPlibData`	In-memory variant of `PlibData`.
`OnDiskPlibData`	Class for handling on-disk data.

Embeddings¶

`list_embeddings`()	Get IDs of registered embeddings.
`describe_embedding`(embedding_id)	Describe specified embedding.
`register_embedding`(embedding_class)	Register new embedding to the collection.
`load_embedding`(embedding_id[, ...])	Load specified embedding as `DataFrame`.

Models¶

Basic routines

`list_models`()	Get IDs of registered models.
`describe_model`(model_id)	Describe specified model.
`register_model`(model_class)	Register new model to the collection.
`load_model`(model_id[, model_args])	Load specified model.
`save_trained_model`(model, path_to_model[, ...])	Load specified model.
`load_trained_model`(path_to_model)	Load model on the specified path.

Base models

`ModelMixin`	Mixin for perturbation models.
`SklearnModel`	Base class for sklearn-style models.

Evaluation¶

Basis routines

`list_evaluators`()	Get IDs of registered evaluators.
`describe_evaluator`(evaluator_id)	Describe specified evaluator.
`register_evaluator`(evaluator_class)	Register new evaluator to the collection.
`load_evaluator`(evaluator_id)	Load specified evaluator.

Mixin

PlibEvaluatorMixin

Mixin for all the Perturb-lib evaluators.

Collections¶

Models

`baselines.NoPerturb`	Replaces any perturbation symbols with a no-perturbation symbol.
`baselines.GlobalMean`	Computes mean value from the training data and then uses it to make predictions.
`baselines.ReadoutMean`	Predicts mean readout value without taking perturbation information into account.
`baselines.Catboost`	CatBoostRegressor used on top of predefined embeddings.
`lpm.LPM`	Large perturbation model.

Evaluators

`standard_ones.RMSE`	Root-mean-square error (RMSE).
`standard_ones.MAE`	Mean absolute error (MAE).
`standard_ones.R2`	R2 score function.
`standard_ones.Pearson`	Pearson correlation coefficient.