Optimal screening for promising genes in 2-stage designs

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Detecting biologically relevant genetic markers is challenging due to multiple testing issues.
• Standard methods prioritize Type I error control, potentially limiting statistical power.
• Existing approaches like the balanced test incorporate the alternative hypothesis for better power.

Purpose of the Study:

• To develop and evaluate two-stage designs for screening genetic markers, particularly when measurement costs are high.
• To optimize decision parameters in two-stage designs to reduce the expected cost per marker.
• To compare the performance of two-stage designs against one-stage designs in terms of cost and accuracy.

Main Methods:

• Utilizing a two-stage sequential sampling approach for genetic marker screening.
• Incorporating the balanced test principle, which considers evidence against both the null and target alternative hypotheses.
• Optimizing decision rules, including the size of the 'gray zone' that triggers second-stage data collection.

Main Results:

• The proposed two-stage designs can substantially reduce the expected cost per marker.
• These designs achieve a superior balance between true negatives and true positives compared to one-stage designs for equivalent costs.
• The method allows for early decisions on markers, gathering more data only when necessary.

Conclusions:

• Two-stage designs offer a cost-effective and statistically powerful alternative for genetic marker screening.
• Optimizing sequential decision-making improves efficiency and the accuracy of genetic marker identification.
• This approach provides a better trade-off between controlling errors and maximizing true positive detection.