A permutation-based non-parametric analysis of CRISPR screen data.

Area of Science:

• Genomics
• Bioinformatics
• Functional Genomics

Background:

• Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) screens are vital for identifying gene functions in cultured cells.
• Existing algorithms for analyzing CRISPR screening data lack standardization and possess limitations.
• There is a need for robust methods to overcome shortcomings in current CRISPR data analysis.

Purpose of the Study:

• To develop and validate a novel algorithm for analyzing CRISPR screening data.
• To provide a more reliable and consistent method for gene function identification.
• To offer an algorithm applicable to various genetic and drug screening datasets.

Main Methods:

• Developed the Permutation-Based Non-Parametric Analysis (PBNPA) algorithm.
• PBNPA computes gene-level p-values by permuting single-guide RNA (sgRNA) labels, avoiding distributional assumptions.
• The algorithm was tested on simulated and published real-world CRISPR, siRNA, shRNA, and drug screening data.

Main Results:

• PBNPA demonstrated superior performance compared to existing methods on both simulated and real data.
• The algorithm showed improved Receiver Operating Characteristics (ROC) curves and False Discovery Rate (FDR) control.
• PBNPA exhibited enhanced consistency and FDR control, particularly with low-quality datasets.

Conclusions:

• The PBNPA algorithm provides more consistent and reliable results for genetic screening data analysis.
• PBNPA is particularly advantageous when dealing with low-quality screening data.
• An R package for PBNPA is publicly available for broader application.