Sample size is a critical component in the design of any high-throughput genetic screening approach. Sample size determination from assumptions or limited data at the planning stages, though standard practice, may at times be unreliable because of the difficulty of a priori modeling of effect sizes and variance. Methods to update the sample size estimate during the course of the study could improve statistical power. In this article, we introduce an approach to estimate the power and update it continuously during the screen. We use this estimate to decide where to sample next to achieve maximum overall statistical power. Finally, in simulations, we demonstrate significant gains in study recall over the naive strategy of equal sample sizes while maintaining the same total number of samples.