Brieflow: an integrated computational pipeline for high-throughput analysis of optical pooled screening data

Area of Science:

• Genomics
• Computational Biology
• Cell Biology

Background:

• Optical pooled screening (OPS) is vital for functional genomics but faces analysis challenges.
• Massive datasets and complex data integration hinder OPS insights.
• Standardized frameworks for OPS data analysis are lacking.

Purpose of the Study:

• To develop Brieflow, an end-to-end computational pipeline for fixed-cell OPS data analysis.
• To introduce MozzareLLM, an AI framework for biological interpretation of OPS data.
• To improve the discovery of genetic functions and cellular phenotypes from high-content screening.

Main Methods:

• Developed Brieflow, a modular, open-source computational pipeline for OPS data.
• Applied Brieflow to reanalyze a large CRISPR-Cas9 screen (5072 genes, >70 million cells).
• Integrated MozzareLLM, a large language model framework, for phenotypic cluster analysis and gene prioritization.

Main Results:

• Brieflow processed massive OPS datasets efficiently, enabling analysis of millions of cells.
• MozzareLLM identified biological processes and prioritized gene candidates for validation.
• The combined approach recovered novel biological modules, including five mitochondrial sub-programs, missed by previous analyses.

Conclusions:

• Brieflow and MozzareLLM provide a robust framework for OPS data analysis and biological interpretation.
• The tools enhance the discovery potential of high-content phenotypic screening.
• The modular and open-source nature promotes reproducibility and further development in functional genomics.