Samuel S. Wang

Office Phone

Research Focus

Information Processing and Learning in Mammalian Brains

The Wang laboratory does research in two areas: (1) contributions of the cerebellum to sensorimotor processing, cognition, and affect?and (2) the postnatal development of autism spectrum disorder. Research interests are represented both by published papers and by more recent work. (For Sam Wang's democracy research see the Princeton Gerrymandering Project as well as published and ongoing projects.)

The cerebellum’s role in learning and action. The cerebellum acts to modify sensation and movement on a subsecond time scale. It does so using highly regular microcircuitry, which despite that regularity partners with a wide variety of brain regions to regulate both motor and nonmotor functions. A major path for cerebellar output passes through thalamus to corticostriatal systems. We are investigating this pathway’s role in flexible behavior, sensory working memory, and social interaction.

In mice we study two kinds of behavior: (a) defined tasks, where we study neural coding principles that underlie sensorimotor learning and working memory; and (b) free behavior, where we track movement and behavioral state. Our findings fit into a framework in which unexpected events, whether sensory, aversive, or rewarding, come in through the climbing fiber pathway to excite Purkinje cell dendrites, thus triggering firing events that act as instructive signals to shape the cells’ output. The instructive signals act on the cerebellum’s ability to process inputs that arrive as a continuous stream via the mossy-fiber/granule-cell pathway, and that depart via deep nuclei and other other pathways.

Current experiments include the use of next-generation calcium sensors to decode granule cell/Purkinje cell activity during associative multisensory learning; silicon-probe recording and optogenetic perturbation during a working memory task; and tracking movement and social interactions to test the effects of perturbing cerebellar activity.

The postnatal progression of autism spectrum disorder. An exciting and rapidly-growing area in autism research is the identification of the neural mechanisms responsible for specific behaviors. We are testing the idea that cerebellar dysfunction during development can produce lasting deficits in sensory processing, social behavior, and cognition. We have found that mice with the same genetic disruptions as those found in autistic persons can disrupt cerebellar learning, and that disruption during development of just the cerebellum can lead to a wide range of autism-like endophenotypes. We are now investigating the possible mechanisms of brainwide influence using silicon-probe recording, whole-brain cellular activity mapping, and in vivo multiphoton imaging of neocortical structural plasticity.

Data analysis methods. Experimental methods in the lab include multiphoton imaging, genetically encodable calcium sensors, viral tracers, brainwide mapping of anatomy and activity, and automated tracking of free behavior. These methods produce large amounts of data which we analyze using machine vision, computational modeling, pose tracking, and latent state analysis. 

Several projects require modeling states that are not directly observable, from molecules to cognitive state. At the molecular level, signals from GCaMP, a popular series of fluorescent calcium indicator proteins, can provide information on action potential timing with 10-millisecond precision. We can reach this level of precision using a generative model that incorporates GCaMP’s hidden dynamics. During an evidence-accumulation decision-making task (see the BRAIN CoGS collaboration), we use latent-state analysis to identify shifts in how evidence is processed. These shifts reveal strategies to learn, stay on task, and perform accurately. Finally, latent-state analysis in solitary and interacting animals can reveal moment-to-moment changes in behavioral state.

Synaptic learning rules. In past years the laboratory has identified fundamental principles by which molecular signaling mechanisms shape learning rules. For example, we found that calcium signaling mechanisms drive the switchlike strengthening and weakening of single synapses. The likelihood and direction of this change is closely dependent on the precise occurrence of certain presynaptic and postsynaptic spike patterns. 

Brain scaling and evolution. In mammals and birds, we have used comparative biophysical methods to infer functional principles of brain architecture. In mammals, we studied the neocortex, whose regularities of structure may be subject to universal design constraints. Whale brains are over 100,000 times larger than shrew brains and are much more folded - yet both species have similar microcircuitry. We found that folding and microcircuitry are related: On average, axons are wider in large brains, and the space demanded by axons is sufficient to account for their increased folding, as well as power-law behavior across species. Widening of axons may be driven by an evolutionary need to preserve fast transmission of nerve impulses across the brain. Since most neural signaling is internal to the brain, we have also investigated the internal proportions of brain components to define a “cerebrotype” whose variations are strongly related to phylogenetic relationships and behavioral complexity.



Related Links

Princeton Neuroscience Institute
Research Area
Systems & Circuits
Molecular & Cellular