Classical dynamical systems are adequately described by the theory of dynamical systems, synthesis of Newton's mechanics and Boltzmann's statistical description, which allowed to found the laws of thermodynamics on a microscopic mechanical description. Living organisms, from the simples macromolecule, capable of self replication, up to homo sapiens, which represents the final point of an evolution lasting over three billion years, are not reducible to dynamical system models. What makes the difference between the simplest living organism and a gas molecule, a ion in a plasma or a spinning top, is its level of organization, to which corresponds a very high information content, its capacity to exchanging information with the environment and of self replicating. The simplest living unit, the complex atom, is far from being representable with a point mass, even with some internal degrees of freedom. The study of living organisms and their aggregates needs a new conceptual and mathematical scheme, we might call theory of complexity, even though this theory does not yet exist in a definite form, universally accepted. Mathematical tools can be found in the information, automata and networks theory, in the games and Darwin's evolution theory, the pillar of theoretical biology. The theory of dynamical systems remains a very useful tool, especially for the construction of phenomenological models. Virtual experiments, which come along with real real ones, may be helpful in finding new metaphors, in detecting new control parameters, and perhaps in finding some general laws, as it happens for physical systems. Mathematical-physics modeling and virtual experiments in the search of simple laws to describe their results are the tools a physicist may use in approachicing biological systems. The activity of our group concerns the neural and genic networks the statistical analysis of genome, the development of dynamical models for proteins and for the immune system. The neural networks and the multiscale analysis, of which we develop some mathematical aspects, are applied to the biomedical images and biological time series analysis. The expression of thousands of genes, experimentally accessible by the use of micro-arrays, requires the use of sophisticated techniques of statistical analysis and dynamical modeling (rivedere testo italiano). The mechanisms which govern the folding of proteins are being considered by proposing classical models of points or rigid bodies chains, to describe single amino-acids. Models for the immune system, based on stochastic population dynamics, have been developed to describe the conversion of T lymphocytes from virgin to memory state and the corresponding clone expansion, by taking into account the antigenic chronic noise. The model allowed to correlate the concentration data with demographic data such as survival curves.