Sequential stopping for high-throughput experiments.

Area of Science:

• Biostatistics
• Genomics
• Proteomics

Background:

• High-throughput experiments often use informal sample size selection.
• Traditional sample size calculations heavily rely on prior knowledge, limiting flexibility.
• Existing methods struggle with the dynamic nature of accumulating data in large-scale studies.

Purpose of the Study:

• To develop a sequential experimental design strategy for high-throughput studies.
• To reduce critical dependence on prior assumptions in sample size determination.
• To provide a coherent and efficient framework for adaptive data acquisition.

Main Methods:

• A sequential strategy that updates knowledge as new data become available.
• A decision-theoretic framework guiding experiment continuation or termination.
• Simulation-based approximation using decision boundaries for practical implementation.
• Application to RNA-seq, microarray, and reverse-phase protein array data.

Main Results:

• The proposed sequential strategy is less critically dependent on initial assumptions.
• Experiments are adaptively stopped or continued based on the utility of additional data.
• The method demonstrates potential advantages in RNA-seq, microarray, and protein array studies.
• The approach is implemented in the Bioconductor package 'gaga'.

Conclusions:

• Sequential adaptive sampling offers a more robust approach to sample size determination in high-throughput experiments.
• The decision-theoretic framework ensures coherent and efficient experimental design.
• This method enhances the practical application of statistical design in genomics and proteomics research.