Searching for meaningful feature interactions with backward-chaining rule induction

Exploring the vast number of possible feature interactions in domains such as gene expression microarray data is an onerous task. We propose Backward-Chaining Rule Induction (BCRI) as a semi-supervised mechanism for biasing the search for plausible feature interactions. BCRI adds to a relatively limited tool-chest of hypothesis generation software, and it can be viewed as an alternative to purely unsupervised association rule learning. We illustrate BCRI by using it to search for gene-to-gene causal mechanisms. Mapping hypothesized gene interactions against a domain theory of prior knowledge offers support and explanations for hypothesized interactions, and suggests gaps in the current domain theory, which induction might help fill.