Research interests

Our research interests are centered on understanding how the neocortex processes sensory information and generates internal models of the world. Specific projects in our group are as following.

Sensory representation in cortical areas

Sensory information from the external environment is processed and transformed into perception and action largely by the neocortex. This process is shaped by experience, as learning a new task can improve and stabilize the intrinsically noisy cortical sensory representation. We previously used a texture discrimination task for mice and uncovered learning-related changes in the primary somatosensory cortex (S1) and posterior parietal areas (PPC). We are interested in further dissecting the function subsets of neural populations and their dynamics underlying sensory representation, and testing various computational models of learning.

Predictive processing in cortical circuits

Our brains are constantly making predictions about the world. The predictive processing hypothesis proposes that the brain generates predictions and continuously compares them with sensory inputs to guide behavior. Strong predictions can modify sensory perception, in extreme cases causing hallucination. To study predictive processing in mice, we previously trained mice to perform a multisensory discrimination task, where sequential auditory and tactile stimuli were paired and allowed mice to form specific sensory predictions. Using this paradigm, we uncovered the dynamic interaction between top-down and bottom-up information in S1 and PPC shapes sensory encoding and affects perceptual choice. We aim to further understand the general signatures of predictive signals in the cortex as well as the interactions between cortical areas underlying sensory-prediction processing.

Cortico-cortical interactions

Cortical areas are not isolated islands – they are structurally and functionally connected to each other. During perception and behavior, cortical areas interact with each other as well as other brain structures to process information and generate actions. With the developmental of high-throughput recording tools, we are now able to record from more and more neurons, across multiple brain areas at the same time. Specifically, our group developed a custom built multi-area two-photon microscope that enables simultaneous recording from two cortical areas, with single neuron resolution from a large population. With this tool, we are investigating the general principles of inter-area communication during sensory processing and behavior.

Memory consolidation across the neocortex

Our internal models of the world is shaped by our individual experiences. These experiences are stored in our brain through a process called memory consolidation. During memory consolidation, new memories are first encoded by the hippocampus, but eventually transferred over time to the neocortex for long-term storage. To study this process, we developed a visually-guided multisensory virtual reality paradigm for head-fixed mice under two-photon microscope. We are interested in understanding how the different cortical areas work together to encode and store different aspects of a memory (auditory, visual, tactile cues). We are also interested in the development memory representation over time as well as signatures of internal models, particularly during offline periods where the mice are not engaged in active tasks.

High-throughput imaging techniques

The recent advances in high-throughput recording techniques have made it possible to monitor the neural activity in large populations during behavior. We are interested in applying these techniques, particularly high-throughput two-photon microscopes, to study the neural dynamics underlying sensory processing and memory. We currently have a custom-built multi-area two-photon microscope based on temporal multiplexing principles. We are seeking to combine this method with electrophyiological techniques, fiber optogenetics, and potentially all-optical approaches for simultaneous two-photon imaging and single-cell optogenetics. We are also interested in taking advantage of the host lab’s expertise and apply multi-fiber imaging for simultaneous recording across even more brain areas.

High-dimensional neural data analysis

As we are dealing with highly complex data from large neuronal populations during animal behavior, often from two areas at the same time, we are constantly working towards better understanding the underlying patterns in such datasets. Towards this goal, we apply various analysis techniques to understand the information encoding, population dynamics, and inter-area interactions in our recorded cortical areas.