Pooling Designs for High-Throughput Biological Experiments.

Abstract

High-throughput experiments provide a fast, automated way to obtain massive amounts of information about biological systems. In several instances, these experiments subject a large number of items (biological samples or entities) individually to identical analysis. The information obtained is often redundant and the measurement systems are error-prone. Pooling designs are multiplexing strategies that test mixtures of items, instead of each item individually, in such a way that every item is tested multiple times. Testing mixtures allows for reduction in measurements, while multiple measurements of each item provide robustness. This thesis achieves three specific advances in applying pooling designs for high-throughput biological experiments.

First, a theoretical framework to study pooling designs is described and a specific pooling strategy called the Shifted Transversal Design (STD) is adopted. Its mathematical properties are investigated and found to be useful but some limitations are also identified. Practical constraints encountered in implementing the STD strategy for a high-throughput drug screening application are investigated and novel solutions are provided.

Second, the results of a proof-of-concept test of the STD strategy in an actual high-throughput screen are described. Robust decoding strategies are investigated and practical solutions are provided. A novel quantitative decoding procedure is developed to exploit the quantitative nature of data obtained from high-throughput screens. The performances of these decoding strategies are described in the context of the proof-of-concept screen.

Third, the quantitative decoding approach is extended to a novel application of pooling designs to gene expression microarray experiments. The results from a small proof-of-concept test of this approach in a microarray experiment are described. Conditions for the successful implementation of this approach are discusses in the context of the experimental results.

Overall, this work introduces and consolidates an approach to biological experimentation that is novel and promises to improve the efficiency and robustness of such experiments. Literature from diverse fields is brought together in the service of advancing the pooling approach and making it practical for existing experimental platforms. Two applications of this approach were investigated experimentally and provided several useful lessons for the success of this approach in these and other applications.