parquet-feature-loading

name	parquet-feature-loading
description	How to load krewlyzer parquet feature files with correct suffixes and schemas per feature type.

Parquet Feature Loading

When to use this skill

• Implementing a new FeatureEvaluator
• Loading sample feature data in exploration cells
• Debugging missing or malformed feature files

Loading Pattern

from kreview.core import load_sample_feature

# Load a feature parquet for one sample
df = load_sample_feature(
    sample_id="P-0000280-T02-XS1",
    feature_suffix=".FSC.gene.parquet",  # exact suffix from §4 table
    results_dir="/path/to/results/",
)

Feature File Suffix Reference

Evaluator	source_file
FSC.gene	.FSC.gene.parquet
FSC bin-level	.FSC.ontarget.parquet
FSC.regions	.FSC.regions.parquet
FSD	.FSD.ontarget.parquet
FSR	.FSR.ontarget.parquet
OCF ontarget	.OCF.ontarget.parquet
OCF offtarget	.OCF.offtarget.parquet
ATAC	.ATAC.ontarget.parquet
TFBS	.TFBS.ontarget.parquet
MDS	.MDS.ontarget.parquet
MDS.gene	.MDS.gene.parquet
MDS.exon	.MDS.exon.parquet
EndMotif	.EndMotif.ontarget.parquet
BreakPointMotif	.BreakPointMotif.ontarget.parquet
EndMotif1mer	.EndMotif1mer.parquet
WPS.panel	.WPS.panel.parquet
WPS genome	.WPS.parquet
WPS background	.WPS_background.parquet
metadata	.metadata.parquet

Anti-Patterns

• ❌ Loading .tsv.gz or .features.json
• ❌ Constructing file paths manually — use load_sample_feature()
• ❌ Loading .correction_factors.parquet or .fsc_counts.parquet (excluded)