Agent skill

bio-crispr-screens-jacks-analysis

JACKS (Joint Analysis of CRISPR/Cas9 Knockout Screens) for modeling sgRNA efficacy and gene essentiality. Use when analyzing multiple CRISPR screens simultaneously or when accounting for variable sgRNA efficiency across experiments.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/jacks-analysis

SKILL.md

JACKS CRISPR Screen Analysis

JACKS jointly models sgRNA efficacy and gene essentiality across multiple experiments. It infers both gene-level fitness effects and sgRNA-specific efficiency.

Installation

bash
pip install jacks
# or
git clone https://github.com/felicityallen/JACKS.git
cd JACKS && pip install -e .

Input File Formats

Count Data

# counts.txt (tab-separated)
sgRNA	Gene	Sample1	Sample2	Sample3	Control1	Control2
sgRNA1	GENE_A	100	120	90	80	85
sgRNA2	GENE_A	200	180	210	150	160
sgRNA3	GENE_B	50	45	55	60	58
...

Replicate Map

# replicatemap.txt
Sample1	Experiment1	Day14
Sample2	Experiment1	Day14
Sample3	Experiment2	Day14
Control1	Experiment1	Day0
Control2	Experiment2	Day0

Guide-Gene Map

# guidemap.txt
sgRNA1	GENE_A
sgRNA2	GENE_A
sgRNA3	GENE_B
sgRNA4	GENE_B
...

Basic JACKS Analysis

Command Line

bash
# Run JACKS
python -m jacks.run_JACKS \
    counts.txt \
    replicatemap.txt \
    guidemap.txt \
    output_prefix \
    --ctrl_sample_pattern "Day0" \
    --ctrl_sample_pattern_column "Condition"

Python API

python
from jacks import infer
import pandas as pd

# Load data
counts = pd.read_csv('counts.txt', sep='\t', index_col=0)
guide_gene_map = pd.read_csv('guidemap.txt', sep='\t', header=None, names=['sgRNA', 'Gene'])
replicate_map = pd.read_csv('replicatemap.txt', sep='\t', header=None,
                             names=['Sample', 'Experiment', 'Condition'])

# Separate control and treatment samples
ctrl_samples = replicate_map[replicate_map['Condition'] == 'Day0']['Sample'].tolist()
treatment_samples = replicate_map[replicate_map['Condition'] == 'Day14']['Sample'].tolist()

# Run JACKS inference
# n_iterations=10000: MCMC iterations. Increase for final analysis.
# burn_in=1000: Burn-in period. Should be ~10% of iterations.
jacks_results = infer.run_inference(
    counts,
    guide_gene_map,
    treatment_samples,
    ctrl_samples,
    n_iterations=10000,
    burn_in=1000
)

Output Files

File Description
_gene_JACKS_results.txt Gene-level essentiality scores
_grna_JACKS_results.txt sgRNA-level efficacy estimates
_jacks_full_data.pickle Full model for downstream analysis

Interpret Gene Results

python
import pandas as pd
import numpy as np

# Load gene results
genes = pd.read_csv('output_gene_JACKS_results.txt', sep='\t')

# JACKS score: negative = essential (dropout), positive = enriched
# Columns: gene, X1 (effect), X2 (std), fdr_log10

# Essential genes (significant negative effect)
# fdr_threshold=-1: log10(FDR) < -1 means FDR < 0.1
essential = genes[(genes['X1'] < 0) & (genes['fdr_log10'] < -1)]
essential = essential.sort_values('X1')
print(f'Essential genes: {len(essential)}')
print(essential.head(20))

# Enriched genes
enriched = genes[(genes['X1'] > 0) & (genes['fdr_log10'] < -1)]
enriched = enriched.sort_values('X1', ascending=False)
print(f'Enriched genes: {len(enriched)}')

sgRNA Efficacy Analysis

python
import pandas as pd

# Load sgRNA results
guides = pd.read_csv('output_grna_JACKS_results.txt', sep='\t')

# Efficacy scores range from 0 (ineffective) to 1 (highly effective)
# X1 column contains efficacy estimates

# Identify poor sgRNAs
# efficacy<0.3: sgRNAs with low efficacy. Consider removal in future libraries.
poor_guides = guides[guides['X1'] < 0.3]
print(f'Low efficacy guides: {len(poor_guides)}')

# Group by gene to assess library quality
gene_efficacy = guides.groupby('Gene')['X1'].agg(['mean', 'std', 'count'])
gene_efficacy = gene_efficacy.sort_values('mean')
print(gene_efficacy.head(20))

Visualization

Gene Effect Plot

python
import matplotlib.pyplot as plt
import numpy as np

genes = pd.read_csv('output_gene_JACKS_results.txt', sep='\t')

fig, ax = plt.subplots(figsize=(10, 8))

# Color by significance
colors = ['red' if fdr < -1 else 'gray' for fdr in genes['fdr_log10']]

ax.scatter(genes['X1'], -genes['fdr_log10'], c=colors, alpha=0.5, s=10)
ax.axhline(1, linestyle='--', color='black', alpha=0.5)  # FDR = 0.1
ax.axvline(0, linestyle='-', color='gray', alpha=0.3)

ax.set_xlabel('JACKS Score (negative = essential)')
ax.set_ylabel('-log10(FDR)')
ax.set_title('JACKS Gene Essentiality')

# Label top hits
top = genes[genes['fdr_log10'] < -2].nsmallest(10, 'X1')
for _, row in top.iterrows():
    ax.annotate(row['gene'], (row['X1'], -row['fdr_log10']))

plt.savefig('jacks_volcano.png', dpi=150)

sgRNA Efficacy Distribution

python
import matplotlib.pyplot as plt

guides = pd.read_csv('output_grna_JACKS_results.txt', sep='\t')

plt.figure(figsize=(8, 5))
plt.hist(guides['X1'], bins=50, edgecolor='black')
plt.axvline(0.5, color='red', linestyle='--', label='Efficacy = 0.5')
plt.xlabel('sgRNA Efficacy')
plt.ylabel('Count')
plt.title('sgRNA Efficacy Distribution')
plt.legend()
plt.savefig('sgrna_efficacy.png', dpi=150)

Multi-Screen Analysis

JACKS strength is joint analysis across experiments.

python
# Define multiple experiments in replicate map
# replicatemap.txt:
# Sample        Experiment    Condition
# Screen1_T1   Screen1       Treatment
# Screen1_T2   Screen1       Treatment
# Screen1_C1   Screen1       Control
# Screen2_T1   Screen2       Treatment
# Screen2_T2   Screen2       Treatment
# Screen2_C1   Screen2       Control

# JACKS will learn shared sgRNA efficacy across screens
# while estimating screen-specific gene effects

Comparing JACKS vs MAGeCK

Feature JACKS MAGeCK
sgRNA efficacy modeling Yes No
Multi-experiment joint analysis Yes Limited
Statistical framework Bayesian MLE/RRA
Speed Slower Faster
Best for Multiple screens Single screen

Advanced Options

python
from jacks import infer

# Run with custom parameters
results = infer.run_inference(
    counts,
    guide_gene_map,
    treatment_samples,
    ctrl_samples,
    n_iterations=50000,     # 50000: Publication quality. 10000 for exploration.
    burn_in=5000,           # 5000: 10% of iterations.
    apply_w_hp=True,        # Hierarchical prior on efficacy
    fixed_w=False,          # Learn sgRNA efficacy (set True to fix at 1)
    w_alpha=0.5,            # Prior shape for efficacy
    w_beta=0.5              # Prior rate for efficacy
)

Integration with Other Tools

Compare with MAGeCK

python
import pandas as pd

jacks = pd.read_csv('jacks_gene_results.txt', sep='\t')
mageck = pd.read_csv('mageck.gene_summary.txt', sep='\t')

# Merge results
merged = pd.merge(jacks, mageck, left_on='gene', right_on='id')

# Compare rankings
from scipy.stats import spearmanr
corr, pval = spearmanr(merged['X1'], merged['neg|score'])
print(f'Spearman correlation: {corr:.3f} (p={pval:.2e})')

Use sgRNA Efficacy for Library Design

python
# Extract high-efficacy guides for future libraries
guides = pd.read_csv('output_grna_JACKS_results.txt', sep='\t')

# efficacy>0.7: High efficacy sgRNAs for optimized libraries.
good_guides = guides[guides['X1'] > 0.7][['sgRNA', 'Gene', 'X1']]
good_guides.to_csv('high_efficacy_guides.csv', index=False)

Related Skills

  • mageck-analysis - Alternative screen analysis method
  • hit-calling - Statistical hit identification
  • screen-qc - Quality control before analysis
  • batch-correction - Handle batch effects in multi-screen data

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results