Variational Bayes approach (GEMINI)

Variational Bayes approach (GEMINI) applies variational Bayesian inference to identify genetic interactions from pairwise combinatorial CRISPR knockout screens while accounting for sample, reagent, and biological variability.

Key Features:

• Variational Bayesian Framework: Models genetic interactions using variational Bayes to produce probabilistic estimates of interaction strengths and to account for experimental variability.
• Joint Analysis Capability: Performs joint analysis of all samples and reagents within an experiment to integrate complex datasets across pairwise knockout designs.
• Scalability and Accuracy: Scales to extensive combinatorial CRISPR screens and improves accuracy in detecting genetic interactions across diverse experimental dimensions.

Scientific Applications:

• Genetic interaction mapping: Systematically identifies pairwise gene–gene interactions from combinatorial CRISPR screens.
• Systems biology: Reveals interaction networks that inform systems-level understanding of biological processes.
• Synthetic lethality discovery: Detects synthetic lethal gene pairs from pairwise knockout data.
• Combinatorial perturbation analysis: Characterizes combinatorial effects of gene knockouts on cellular phenotypes using integrated reagent and sample data.

Methodology:

GEMINI applies variational inference to model genetic interactions and probabilistically estimate interaction strengths from pairwise CRISPR screen data, leveraging all available sample and reagent observations to mitigate experimental noise and variability.