The way we think, what we learn and remember, and how we sense, feel, and act, are all engraved in the neuronal connections of our brain. We are alike and we can communicate because we share common structures in our neural networks. We have different habit, talent, and personality depending on the diversity in detailed structure of the neural networks. As we learn new things and store them in the brain, the neural network is re-wired to change its synaptic connections. The entire neuronal connectivity that inherently represents everything of us is called the Connectome, or the Brain Map. Professor Sebastian Seung aphoristically said in his book and TED talk that "I am my connectome."

These are strong but yet to be proven hypotheses. Computational Neuroscience is a study based on the assumption that the brain is a biological computer that works under such hypotheses. To truly understand a computer (brain), we need to understand the functions of its logic gates (neurons) and their wiring (neuronal circuits) that yields higher level functions, in a bottom-up manner. Historically, computational neuroscience has striven to suggest mathematical models of the neural computations based on biological and physiological experiments and thus to produce theoretical foundations for understanding the brain functions — firstly for individual neurons and then for neuronal circuits. Connectomics inherits and realizes these ideas by considering the ultrastructural connection specificity and the complete population of neurons in circuits — firstly for small samples or regions of the brain and ultimately for the entire brain. For its success, it is crucial to integrate the studies from neuroanatomy, electrophysiology, and molecular biology.

In our lab we observe the 3D structure of neurons and their connectivity from high-resolution images obtained by serial electron and light microscopes. We use deep learning artificial intelligence and other computational techniques to analyze the images. They become anatomical component of the connectome. We also computationally analyze the neural activity data such as calcium imaging and electrophysiological recordings. They become physiological component of the connectome. We mathematically model the functions of neural networks from the activity of neurons in the circuit. We collaborate with the labs at Korea Brain Research Institute (KBRI) and other institutes for the acquisition of experiment data.

One of our current projects is a study on the cerebellum. To study the computational mechanism of cerebellar motor control and learning, we applied our serial electron microscope (EM) image reconstruction pipeline to a small patch of mouse cerebellar molecular layer. Up to our knowledge, this is the first of a whole-cell scale EM reconstruction of the Purkinje cells. Moreover, we are discovering interesting subcellular wiring specificity between the Purkinje cells and their input neurons. The gallery below showcases our reconstructed cerebellar neural circuits. Click the figure to find out more.