scripts/6.feature-select.py
Performs feature selection on normalized profiles with Pycytominer (by default: variance threshold, correlation threshold, dropping NA columns, blocklist, and dropping outliers).
group sets which profiles are pooled to decide which features to keep: each plate, all plates in a data set (group), or all data sets together.
Inputs¶
| Input | Description |
|---|---|
profiles/{plate}_{data_set_name}_{gene|guide}_normalized.csv.gz | From step 5 |
Config used: data_sets (experiment); feature_select.*, core.* (defaults). See Configuration.
Outputs¶
| Output | Description |
|---|---|
profiles/{plate}_{data_set_name}_{gene|guide}_normalized_feature_selected_by{group}.csv.gz | Feature selected profiles; {group} is the feature_select: group value (plate, group, or all) |