Upinder Bhalla
Keynote lecture
Will talk about: Multiscale models of the synapse: a self-modifying memory machine
I was interested in physics and computers through school and college, but the biology undergraduate courses at Cambridge captivated me and I decided to switch fields for my PhD at Caltech. Among the most inspiring moments of my bachelor's courses were the realization that bacteriophage life-cycles could actually be thought of as hacked- together living programs, and the glimpse of computation happening in the brain. I took these interests to my PhD, which was on an experimental and computational study of the rat olfactory system. My post-doc research at Mount Sinai was also serendipitous, starting with the idea that perhaps molecular interactions in the neuron could bbe modeled computationally, and eventually leading into one of the early kinetic models of signaling pathways in memory storage. Since 1996 I have been at the National Centre for Biological Sciences in Bangalore, and I have continued to pursue both experiments and models in my research into neural computation at levels ranging from single molecules up to the entire brain.
The human brain expresses some 20,000 genes, 100 billion neurons, and around 10 to the power of 15 synapses that connect up the neurons. Purely on a numerical basis, it seems likely that the synaptic connections would be a good place to store the vast amount of information that makes up our memories. There is now a considerable body of experimental data to show that synapses change in an experience-dependent manner, and increasingly point to these modifications as a key cellular basis for memory. This turns out to be a fertile and challenging arena for multiscale modeling and neuroinformatics. Synapses are precisely at the junction of electrical and chemical signaling. Although there are a plethora of models of signaling in memory, they are small pieces in a multidimensional puzzle. Synaptic memory is one of those processes which demand not just signaling models, but multiscale models that encompass neuronal networks, cellular biophysics, structural change, biochemical signaling, protein synthesis, and gene expression. Some of these domains - like biochmical signaling - are well-represented by simulation tools and standards such as SBML. Others - like structural change - have few, mostly incompatible, tools. I will present the process of modeling the synapse across a few of these multiple scales. There are conceptual challenges here, since we are fundamentally trying to understand how to get immensely stable, life-long changes out of a system that can not only reprogram itself, but also rebuild itself. Other challenges are to see how the synapse balances the requirements for fast switching, against long-term stability in the face of biochemical stochasticity. There are interesting couplings across scales, where electrical events have biochemical effects, and vice versa. I suggest that this cross-coupling at the synapse is one of the key systems where the convergence of neuroinformatics tools and standards can make a huge difference.
