Group testing provides a powerful way to get the most out of each experiment, but group testing relies on optimized d-disjunct matrices and t-designs. While there are fast ways to generate these designs, in some cases the best designs can only currently be found by random search.
An important repostory of results are translated from Dan Gordon's collection of covering designs
. Others are pulled from the literature or generated internally.
A curated subset of these designs are presented in the Origami Assays Project
. That site also includes further documentation and decoders.
Feel free to browse the designs
This page stores and shares hundreds of optimized designs with the following properties:
Disjunct matrix: Each design is a matrix of assays, and each assay contains a list of numbered sample IDs. The underlying structure of this matrix must be disjunct.
Equal coverage: Each sample must be replicated the same number of times in the design. This replication number is one greater than the disjunct level.
Include between 2 and 21 replicates (r): We are constraining our search to disjunct designs of order between 1 and 6. Designs with more replicates are possible, but they provide no compression with under 100 tests.
Include between 10 and 100 tests (m): We are constraining the dictionary to a limited number of tests because these are the regions that are most experimentally tractable.
Optimized: The best design for a given number of tests (m or v) and replicates (r or k) will assay the largest number of samples possible. If a better design is found (greater n for a given m and r), it will take the lead position.