drcme.bin.run_rf_prediction

Script to predict type labels for new data using a random forest classifier and training data.

The electrophysiology files are split such that the reference_ephys_file contains data for training the classifier and prediction_ephys_file contains data for new predictions. The morph_file, if used, has morphology data for both sets of cells.

class drcme.bin.run_rf_prediction.RfPredictionParameters(extra=None, only=None, exclude=(), prefix='', strict=None, many=False, context=None, load_only=(), dump_only=(), partial=False)

Parameter schema for random-forest prediction

This schema is designed to be a schema_type for an ArgSchemaParser object

RfPredictionParameters

key	description	default	field_type	json_type
input_json	file path of input json file	NA	InputFile	str
output_json	file path to output json file	NA	OutputFile	str
log_level	set the logging level of the module	ERROR	LogLevel	str
reference_ephys_file	Path to electrophysiology data file for reference cells	NA	InputFile	str
prediction_ephys_file	Path to electrophysiology data file for cells that will have predicted labels	NA	InputFile	str
reference_label_file	Path to type labels for reference cells	NA	InputFile	str
label_key	Column name of type label in ‘reference_label_file’	NA	String	str
morph_file	Path to morphology data file for all cells	None	InputFile	str
output_file	Path to output file with predicted labels	NA	OutputFile	str
ref_id_file	Path to file with subset of IDs for reference cells	None	InputFile	str
pred_id_file	Path to file with subset of IDs for predicted cells	None	InputFile	str
n_trees	Number of trees for random forest classifier	500	Integer	int
class_weight	Class weight parameter for random forest classifier	None	String	str

drcme.bin.run_rf_prediction.construct_datasets(ephys_ref, ephys_pred, morph_df)

Build reference and test data sets

Parameters

• ephys_ref (DataFrame) – DataFrame with reference electrophysiology data

• ephys_pred (DataFrame) – DataFrame with electrophysiology data for label prediction

• morph_df (DataFrame) – DataFrame with morphology data for all cells

Returns

• ref_df (DataFrame) – Combined ephys/morph data set for reference cells

• test_df (DataFrame) – Combined ephys/morph data set for cells that will have labels predicted

drcme.bin.run_rf_prediction.intersect_ephys_morph(ephys_df, morph_df)

Make combined DataFrame with shared cells from ephys_df and morph_df

Parameters

• ephys_df (DataFrame) – DataFrame with electrophysiology data

• morph_df (DataFrame) – DataFrame with morphology data

Returns

Combined ephys/morph data set

Return type

DataFrame

Functions

construct_datasets(ephys_ref, ephys_pred, …)	Build reference and test data sets
intersect_ephys_morph(ephys_df, morph_df)	Make combined DataFrame with shared cells from ephys_df and morph_df
main(reference_ephys_file, …)	Main runner function for script.

Classes

RfPredictionParameters([extra, only, …])

Parameter schema for random-forest prediction