A Comparative Survey of Non-Adaptive Pooling Designs

Abstract

Pooling (or “group testing”) designs for screening clone libraries for rare “positives” are described and compared. We focus on non-adaptive designs in which, in order both to facilitate automation and to minimize the total number of pools required in multiple screenings, all the pools are specified in advance of the experiments. The designs considered include deterministic designs, such as set-packing designs, the widely-used “row and column” designs and the more general “transversal” designs, as well as random designs such as “random incidence” and “random k- set” designs. A range of possible performance measures is considered, including the expected numbers of unresolved positive and negative clones, and the probability of a one-pass solution. We describe a flexible strategy in which the experimenter chooses a compromise between the random k -set and the set-packing designs. In general, the latter have superior performance while the former are nearly as efficient and are easier to construct.