FEAT_PATH = Path(SAVE_DIR) / 'features.csv'
if not META_PATH.exists() and not FEAT_PATH.exists():
# compile metadata
metadata, metadata_path = compile_metadata(aux_dir=AUX_DIR,
imaging_dates=IMAGING_DATES,
n_wells=N_WELLS,
add_well_annotations=N_WELLS==96,
from_source_plate=False)
# compile window summaries
features, metadata = process_feature_summaries(metadata_path=metadata_path,
results_dir=RES_DIR,
compile_day_summaries=True,
imaging_dates=IMAGING_DATES,
align_bluelight=False,
window_summaries=True,
n_wells=N_WELLS)
# clean results
features, metadata = clean_summary_results(features,
metadata,
feature_columns=None,
nan_threshold_row=NAN_THRESHOLD_ROW,
nan_threshold_col=NAN_THRESHOLD_COL,
max_value_cap=1e15,
imputeNaN=True,
min_nskel_per_video=MIN_NSKEL_PER_VIDEO,
min_nskel_sum=MIN_NSKEL_SUM,
drop_size_related_feats=False,
args = load_json(args.json)
aux_dir = Path(args.project_dir) / 'AuxiliaryFiles'
results_dir = Path(args.project_dir) / 'Results'
# load metadata
metadata, metadata_path = process_metadata(
aux_dir,
imaging_dates=args.dates,
add_well_annotations=args.add_well_annotations,
n_wells=6)
features, metadata = process_feature_summaries(metadata_path,
results_dir,
compile_day_summaries=False,
imaging_dates=args.dates,
align_bluelight=False,
window_summaries=True,
n_wells=6)
# Subset results (rows) to remove entries for wells with unknown strain data for 'gene_name'
n = metadata.shape[0]
metadata = metadata.loc[~metadata['gene_name'].isna(), :]
features = features.reindex(metadata.index)
print("%d entries removed with no gene name metadata" %
(n - metadata.shape[0]))
# update gene names for mutant strains
metadata['gene_name'] = [
args.control_dict['gene_name'] if s == 'BW' else s
for s in metadata['gene_name']
TEST_NAME = args.test # str, Choose between 'LMM' (if >1 day replicate), 'ANOVA' or 'Kruskal'
# Kruskal tests are performed instead of ANOVA if check_normal and data is not normally distributed)
# If significant features are found, pairwise t-tests are performed
#%% Compile and clean results
# Process metadata
metadata, metadata_path = process_metadata(aux_dir=AUX_DIR,
imaging_dates=args.dates,
add_well_annotations=args.add_well_annotations)
# Process feature summary results
features, metadata = process_feature_summaries(metadata_path,
RESULTS_DIR,
compile_day_summaries=args.compile_day_summaries,
imaging_dates=args.dates,
align_bluelight=args.align_bluelight)
# Clean: remove data with too many NaNs/zero std and impute remaining NaNs
features, metadata = clean_summary_results(features,
metadata,
feature_columns=None,
imputeNaN=args.impute_nans,
nan_threshold=args.nan_threshold,
max_value_cap=args.max_value_cap,
drop_size_related_feats=args.drop_size_features,
norm_feats_only=args.norm_features_only,
percentile_to_use=args.percentile_to_use)
# Load supplementary info + append to metadata
parser = argparse.ArgumentParser(description="Compile Keio screen metadata & feature summaries")
parser.add_argument('-j', '--json', help="Path to JSON parameters file for analysis",
default=JSON_PARAMETERS_PATH, type=str)
args = parser.parse_args()
args = load_json(args.json)
assert args.project_dir is not None
aux_dir = Path(args.project_dir) / "AuxiliaryFiles"
results_dir = Path(args.project_dir) / "Results"
metadata, metadata_path = process_metadata(aux_dir=aux_dir,
imaging_dates=args.dates,
add_well_annotations=True)
metadata, features = process_feature_summaries(METADATA_PATH, results_dir,
compile_day_summaries=False,
imaging_dates=args.dates,
align_bluelight=False)
features, metadata = compile_bluelight_window_results(args)
# Save features to file
features_path = Path(args.save_dir) / 'features.csv'
features.to_csv(features_path, index=False)
# Save metadata to file
metadata_path = Path(args.save_dir) / 'metadata.csv'
metadata.to_csv(metadata_path, index=False)
toc = time()
print("\nDone in %.1f seconds (%.1f minutes)" % (toc - tic, (toc - tic) / 60))