DISCOVERY BY MINIMAL LENGTH ENCODING - A CASE-STUDY IN MOLECULAR EVOLUTION

We apply the Minimal Length Encoding Principle to formalize inference about the evolution of macromolecular sequences. The Principle is show n to imply a combination of Weighted Parsimony and Compatibility metho ds that have long been used by biologists because of their good practi cal performance. The background assumptions are expressed as an encodi ng scheme for the observed data and as heuristic rules for selection o f diagnostic positions in the sequences. The Principle was applied to discover new subfamilies of Alu sequences, the most numerous family of repetitive DNA sequences in the human genome.