A clear link between topological structure and biological function did emerge in terms of class membership prediction (average 67% of correct predictions, p < 0.0001) with a varying degree of ‘peculiarity’ across classes going from a very low (25%) recognition efficiency for neural and brain networks to the extremely high (80%) peculiarity of amino acid–amino acid interaction (AAI) networks.
We recognized four main dimensions (principal components) as main organization principles of biological networks. These components allowed for an efficient description of network architectures and for the identification of ‘not-physiological’ (in this case cancer metabolic networks acting as test set) wiring patterns.
We highlighted as well the need of developing new theoretical generative models for biological networks overcoming the limitations of present mathematical graph idealizations.