domains / families / science / perturb-seq

perturb-seq

separate

Analyze Perturb-seq / CROP-seq single-cell CRISPR screens. Use when assigning guides as a mixture problem, removing non-perturbed escaper cells with Mixscape, choosing a calibrated test (SCEPTRE conditional resampling) over naive DE, quantifying effect size with E-distance, separating compositional shifts from within-state expression change, or judging whether a perturbation-prediction foundation model actually beats a baseline.

When to use

• Assigning guides as a mixture problem
• Removing non-perturbed escaper cells with Mixscape
• Choosing a calibrated test (SCEPTRE conditional resampling)
• Quantifying effect size with E-distance
• Separating compositional shifts from within-state expression change
• Judging whether a perturbation-prediction foundation model beats a baseline

When not to use

• Naive DE testing
• Tasks not involving single-cell CRISPR screens

Skill page

Source