Toward a Computational Microscope for Neurobiology

Though great strides have been made in neurobiology, we do not yet understand how the symphony of communication among neurons leads to rich, competent behaviors in animals. How do local interactions among neurons coalesce into the behavioral dynamics of nervous systems, giving animals their impressive abilities to sense, learn, decide, and act in the world? Many details remain cloaked in mystery. We are excited about the promise of gaining new insights by applying computational methods, in particular machine learning and inference procedures, to generate explanatory models from data about the activities of populations of neurons. For most of the history of electrophysiology, neurobiologists have monitored the membrane properties of neurons of vertebrates and invertebrates by using glass micropipettes filled with a conducting solution. Mastering techniques that would impress the most expert of watchmakers, neuroscientists have fabricated glass electrodes with tips that are often less than a micron in diameter, and they have employed special machinery to punch the tips into the cell bodies of single neurons—with the hope that the neurons will function as they normally do within larger assemblies. Such an approach has provided data about the membrane voltages and action