Coordinator: Riccardo Zecchina
The research interest of our group lies at the interface between statistical physics, computer science, information theory and computational biology. Its main themes include equilibrium and out-of-equilibrium phenomena in disordered systems, statistical physics of inverse problems, statistical inference and combinatorial optimization, probabilistic message-passing algorithms, stochastic processes, randomized algorithms, graphical games and coding. Currently we are focused on the understanding and the development of probabilistic message passing algorithms (MPA) of different degrees of complexity.
MPA have been developed in the last few years in the context of Statistical Physics of Constraint Satisfaction Problems where they have proven to be effective in dealing with optimization problems over random structures. In many cases, MPA are able to explore efficiently the space of solutions in parallel, maintaining a fully distributed updating scheme that keeps the computational effort at a low level.
A main field of applications of our techniques consists in large scale inverse problems in Computational Biology.
The members of the group cover a wide spectrum of competence, from statistical physics, to computer science and information theory, to computational biology and computational neuroscience.
The research activity in computational biology is supported by Microsoft External Research Initiative.
Nowadays, in order to make innovations in an advanced scientific and technological context - new materials, nanoscience, system biology, neuroscience, computation, network engineering, web economy, financial markets modeling - it is mandatory to master the most advanced concepts and methodologies for complex systems. This is why we have set up a master program, based in prestigious sites in Italy and France, which aims to offer the best techniques needed to attack interdisciplinary problems.
The aim of the international master in Physics of Complex Systems is to shape professionals and/or potential researchers able to jointly apply knowledge and methodologies from modern physics, applied mathematics, information engineering and computational biology to the analysis, modeling and simulation of complex systems.
Direct-coupling analysis of residue coevolution captures native contacts across many protein families
Direct-coupling analysis of residue coevolution captures native contacts across many protein families. Proceedings of the National Academy of Sciences. 2011;108:E1293-E1301.
Protein 3D Structure Computed from Evolutionary Sequence Variation. PLoS ONE. 2011;6:e28766.
A Conference on Statistical physics of complexity, optimization, and systems biology
Our algorithm for the Prize-Collecting Steiner Trees on graphs found better upper bounds for many unsolved instances on the Steinlib Library. The improved costs are: