Translating Time-Course Gene Expression Profiles into Semi-algebraic Hybrid Automata Via Dimensionality Reduction

Biotechnological innovations which sample gene expressions allow to measure the gene expression levels of a biological system with varying degree of accuracy, cost and speed. By repeating the measurement steps at different sampling rates, one can both infer relations among the genes and define a dynamic model of the underlying biological system. When a very large number of genes and measurements are involved, they raise several difficult algorithmic questions, as accurate model-building, checking and inference tasks. Semi-algebraic hybrid automata were proposed as a modeling formalism for biological systems (see, e.g., [17,6]), and demonstrated their abilities to handle complex biochemical pathways. This paper proposes an automatic procedure to build semi-algebraic hybrid automata from gene-expression profiles. In order to reduce the size of the resulting automata and to minimize their analysis computational complexity, our approach exploits various dimensionality reduction techniques. The paper concludes with several experimental results about peach fruit