01/12/11

Neuroscience - Computation with dynamic neural activity patterns

The main function of the nervous system is to process information. There are several models that try to describe how neural information processing works. I am particularly interested in the analysis of spatio-temporal neural activity patterns, and the neural information processing models involving such patterns. Neural systems, which allow the direct study of such patterns are for example the olfactory bulb and the stomato-gastric ganglion of crabs. In the context of neural information processing I am also interested in the evolution of the nervous system and it's information processing abilities.

My current particular interest is the analysis and modelling of the crab stomatogastric ganglion (STG) using high-resolution and high-speed voltage sensitive dye imaging and computational modelling of neurons. This small (26 neurons) neural system with complex behaviour provides an excellent opportunity for the study and understanding of how activity patterns of neurons are used to perform system level computations. The STG being a key model of motor central pattern generator (CPG) neural circuits implies that results obtained in this system can have wide ranging impact in many fields related to study of movement generation and control - for example, restoration of regular activity patterns in de-centralized motor control ganglia with potential impact in spinal cord research, or dopamine modulation of the coordination of movement control neural circuits with potential impact in Parkinson's disease research.

Complex systems - Network analysis, systems theory and modelling

Many complex systems (e.g. neural systems, cells, large-scale software) are organised as networks of interacting components. In complex systems there are many (thousands or more) components and a comparably large number of interactions between these components. For example, a bacterial cell may be able to produce a few thousand proteins that are characterised by a few thousand interactions, or a large-scale software system may have a few thousand classes with a few thousand method calls linking these classes together to deliver the functionality of the software. Understanding such complex networked systems is a major challenge. My interest is in particular in developing and validating network analysis methods that can discover functionally important parts of such networks. Currently my focus is on using large-scale software as a test bed for developing such methods and to prove their usefulness by showing how they can help software engineers to fix and improve the functionality of such software. I also applied these methods to neural systems and protein interaction system. In the latter case the aim was to help the discovery of novel antibiotic targets in bacteria.

Complex systems organize themselves, and their self-organization principles are similar irrespective of their actual context. Understanding such principles and their application in particular contexts may help very much the understanding of how complex systems work and evolve. Systems theory offers many conceptual tools for this work. I am specially interested in how social communication systems organize themselves and evolve, and in how knowledge about the evolution and organization of complex social systems can be applied in the context of other complex communication systems (e.g., neural systems).

Cooperation is a somewhat paradoxical phenomenon in communities of selfish individuals. Game theory and simulation studies provide ways to analyze the emergence of cooperative behavior in communities. In addition, information and systems theoretic consideration may help us to see more clearly the functional integration of components within large systems (e.g., ecosystem), which may determine the effective units of evolution (e.g., individuals or cooperative structures). I am interested particularly in simulation studies and the information and systems theoretic tools of analysis.

Computational intelligence

In the context of applied computational intelligence I am interested in the application of various types of neural networks (e.g., SVMs, Kohonen networks, etc.), evolutionary algorithms, fuzzy decision systems and Bayesian networks to solve real world problems. My relevant interests are medical applications, computational finance, decision support systems, and computational drug design. For example, how can we use machine learning methods to help the objective diagnosis and disease progression evaluation in Parkinson's disease, or how can support the participation of group members in collective decision making using interactive touch sensitive tabletop devices.

The amount of accessible neuroscience data is huge and this data is growing at a very high rate. Besides of the benefits of this it creates also problems related to how to handle the extremely large amount of information in an efficient way. Neuroinformatics research aims to solve such problems by building software solutions that help their user to find the required information to support their decisions (scientific or clinical). In this respect I am interested in developing software tools that can handle large, growing and dynamically changing data in an efficient manner, using specialist neuroscience knowledge. The same problems apply in various other fields of life sciences, and I am also interested in applying e-science solutions in these domains as well (e.g. in intracellular microscopy and mitochondrial bioinformatics).