Machine Learning Bio-molecular Interactions from Temporal Logic Properties


With the advent of formal languages for modeling bio-molecu-lar interaction systems, the design of automated reasoning tools to assist the biologist becomes possible. The biochemical abstract machine BIOCHAM software environment offers a rule-based language to model bio-molecular interactions and an original temporal logic based language to formalize the biological properties of the system. Building on these two formal languages, machine learning techniques can be used to infer new molecular interaction rules from temporal properties. In this context, the aim is to semi-automatically correct or complete models from observed biological properties of the system. Machine learning from temporal logic formulae is quite new however, both from the machine learning perspective and from the Systems Biology perspective. In this paper we present an ad-hoc enumerative method for structural learning from temporal properties and report on the evaluation of this method on formal biological models of the literature.

In Gordon Plotkin, editor, Third Workshop on Computational Methods in Systems Biology