Conditional resampling improves calibration and sensitivity in single cell CRISPR screen analysis

Single-cell CRISPR screens are the most promising biotechnology for mapping regulatory elements to their target genes at genome-wide scale. However, the analysis of these screens presents significant statistical challenges. For example, technical factors like sequencing depth impact not only expression measurement but also perturbation detection, creating a confounding effect. We demonstrate on two recent high multiplicity of infection single-cell CRISPR screens how these challenges cause calibration issues among existing analysis methods. To address these challenges, we propose SCEPTRE: analysis of single-cell perturbation screens via conditional re-sampling. This methodology, designed to avoid calibration issues due to technical confounders and expression model misspecification, infers associations between perturbations and expression by resampling the former according to a working model for perturbation detection probability in each cell. SCEPTRE demonstrates excellent calibration and sensitivity on the CRISPR screen data and yields hundreds of new regulatory relationships, supported by orthogonal functional evidence.