bio-workflows-crispr-screen-pipeline

Step 2: Quality Control

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

counts = pd.read_csv('experiment.count.txt', sep='\t', index_col=0)
counts_numeric = counts.iloc[:, 1:]

qc_stats = {}
for col in counts_numeric.columns:
    total = counts_numeric[col].sum()
    zeros = (counts_numeric[col] == 0).sum()
    gini = calculate_gini(counts_numeric[col].values)
    qc_stats[col] = {'total_reads': total, 'zero_count_guides': zeros, 'gini': gini}

qc_df = pd.DataFrame(qc_stats).T
print('QC Summary:')
print(qc_df)

## QC thresholds

assert qc_df['zero_count_guides'].max() < len(counts) * 0.2, 'Too many zero-count guides'
assert qc_df['gini'].max() < 0.4, 'Gini index too high (uneven distribution)'
print('QC passed!')

For dropout/negative selection screens

mageck test
-k experiment.count.txt
-t Day14_Rep1,Day14_Rep2,Day14_Rep3
-c Day0
-n negative_screen
--pdf-report
--gene-lfc-method alphamedian