ageas.Unit

class ageas.Unit(database_info=None, meta=None, model_config=None, pseudo_grns=None, clf_keep_ratio: float = 0.5, correlation_thread: float = 0.2, cpu_mode: bool = False, feature_dropout_ratio: float = 0.1, feature_select_iteration: int = 3, grp_changing_thread: float = 0.05, max_train_size: float = 0.95, model_select_iteration: int = 2, outlier_thread: float = 3.0, regulatory_trace_depth: int = 1, stabilize_patient: int = 3, stabilize_iteration: int = 10, top_grp_amount: int = 100, z_score_extract_thread: float = 0.0)

Extractor Unit object to get candidate key regulatory pathways and corresponding genes.

Results are stored in attributes and can be saved as files.

__init__(database_info=None, meta=None, model_config=None, pseudo_grns=None, clf_keep_ratio: float = 0.5, correlation_thread: float = 0.2, cpu_mode: bool = False, feature_dropout_ratio: float = 0.1, feature_select_iteration: int = 3, grp_changing_thread: float = 0.05, max_train_size: float = 0.95, model_select_iteration: int = 2, outlier_thread: float = 3.0, regulatory_trace_depth: int = 1, stabilize_patient: int = 3, stabilize_iteration: int = 10, top_grp_amount: int = 100, z_score_extract_thread: float = 0.0)

Initialize a new AGEAS Extractor Unit.

Parameters

• database_info – <object Default = None> Integrated database information returned by ageas.Data_Preprocess()

• meta – <object Default = None> Meta level processed GRN information returned by ageas.Data_Preprocess()

• model_config – <dict Default = None> Dictionary containing configs of all candidate classification models.

• pseudo_grns – <object Default = None> pseudo-sample GRNs returned by ageas.Data_Preprocess()

• clf_keep_ratio – <float Default = 0.5> Portion of classifiers to keep on last model selection iteration.

• correlation_thread –

  <float Default = 0.2> Gene expression correlation thread value of GRPs.

  Potential GRPs failed to reach this value will be dropped.

• cpu_mode – <bool Default = False> Whether force to use CPU only or not. By default, AGEAS will automatically select device favoring CUDA based GPUs.

• feature_dropout_ratio – <float Default = 0.1> Portion of features(GRPs) to be dropped out after each iteration of feature selection.

• feature_select_iteration – <int Default = 3> Number of iteration for feature(GRP) selection before key GRP extraction

• top_grp_amount –

  <int Default = 100> Amount of GRPs an AGEAS extractor unit would extract.

  If outlier_thread is set, since outlier GRPs are extracted during feature selection part and will also be considered as key GRPs, actual amount of key GRPs would be greater.

• grp_changing_thread – <float Default = 0.05> If changing portion of key GRPs extracted by AGEAS unit from two stabilize iterations lower than this thread, these two iterations will be considered as having consistent result.

• model_select_iteration – <int Default = 3> Number of iteration for classification model selection.

• outlier_thread – <float Default = 3.0> The lower bound of Z-score scaled importance value to consider a GRP as outlier need to be retain.

• regulatory_trace_depth – <int Default = 1> Trace regulatory upstream of regulatory sources included in key networks extracted.

• stabilize_patient – <int Default = 3> If stabilize iterations continuously having consistent result for this value, an early stop on result stabilization will be executed.

• stabilize_iteration – <int Default = 10> Number of iteration for a AGEAS unit to repeat key GRP extraction after model and feature selections in order to find key GRPs consistently being important.

• max_train_size –

  <float Default = 0.95> The largest portion of avaliable data can be used to train models.

  At the mandatory model filter, this portion of data will be given to each model to train.

• z_score_extract_thread – <float Default = 0.0> The lower bound of Z-score scaled importance value to extract a GRP.

Methods

__init__([database_info, meta, ...])	Initialize a new AGEAS Extractor Unit.
generate_atlas()	Function to build networks with extracted GRPs.
launch()	Function to launch extractor unit after model selection.
select_models()	Function to perform model selection.