Summary: Inductive Inference over MacroMolecules.
L.Allison, C.S.Wallace and C.N.Yee.
Department of Computer Science,
November 14, 1990
uucp: firstname.lastname@example.org xxx=[lloyd, csw, cyee]
Supported by Australian Research Council grant A48830856
A preliminary version of this paper was presented at a joint session of the AAAI Spring Symposia on
Artificial Intelligence and Molecular Biology and on the Theory and Application of Minimal Length
Encoding, at Stanford, March 2729 1990.
A fuller version titled `FiniteState Models in the Alignment of Macromolecules' later appeared in
J. Mol. Evol. 35 77-89 1992.
Abstract. Minimum message length techniques are applied to problems over strings such as biological
macromolecules. This allows the posterior oddsratio of two theories or hypotheses about strings to be
calculated. Given two strings we compare the rtheory, that they are related, with the nulltheory, that they
are not related. This is done for one, three and fivestate models of relation. Models themselves can be
compared and this is done on real DNA strings and artificial data. A fast, approximate MML string
comparison algorithm is also described.
Keywords. alignment, edit distance, homology, inductive inference, MML, similarity, string.