STAREG: Statistical replicability analysis of high throughput experiments with applications to spatial transcriptomic...

Area of Science:

• Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• Replicable signals across studies strengthen scientific evidence.
• High-throughput experiments generate vast amounts of data, necessitating robust analysis methods.
• Spatial transcriptomic studies offer insights into tissue organization and cellular function.

Purpose of the Study:

• Introduce STAREG, a novel statistical method for replicability analysis in high-throughput experiments.
• Apply STAREG to spatial transcriptomic datasets to identify robust biological signals.
• Enhance statistical power and control the false discovery rate (FDR) in multi-study analyses.

Main Methods:

• STAREG models the joint distribution of p-values from multiple studies, accounting for heterogeneity.
• Utilizes summary statistics from high-throughput experiments.
• Employs the Expectation-Maximization (EM) algorithm and pool-adjacent-violator-algorithm (PAVA) for scalability.

Main Results:

• STAREG effectively controls the false discovery rate (FDR) while increasing statistical power through information borrowing.
• Identifies unique gene rankings, highlighting important genes missed by existing methods.
• Enabled novel biological discoveries in spatial transcriptomic data analysis.

Conclusions:

• STAREG provides a powerful and scalable approach for replicability analysis in high-throughput experiments.
• The method facilitates robust biological discoveries from spatial transcriptomic data.
• STAREG's ability to borrow information across studies improves inference and controls error rates.