Off-manifold coding in visual cortex revealed by sleep
The primary visual cortex (V1) exhibits low-dimensional movement dynamics and high-dimensional sensory representations. How are these contradictory regimes organized in the same neural population? By using activity during non-REM sleep, we revealed a low-dimensional intrinsic manifold structure that preferentially encodes movements. Surprisingly, visual scenes engage an off-manifold subspace, comprising the sparse activity of neurons that are strongly coupled by the low-dimensional intrinsic manifold. These findings highlight how the brain can efficiently use the neural activity space by identifying a key link: low-dimensional manifolds open a sparse coding space for high-dimensional representations. bioRxiv
Identifying patterns differing between high-dimensional datasets with generalized contrastive PCA
Comparing high-dimensional datasets is common in biology, but identifying contrastive low-dimensional patterns remains challenging. Traditional dimensionality reduction methods process single datasets at a time, and existing tools are limited by the need for hyperparameter tuning and asymmetric data handling. We developed generalized contrastive PCA (gcPCA), a hyperparameter-free method that symmetrically compares datasets and provides a mathematically rigorous framework to contrast high-dimensional data. PLOS Computational Biology
Time encoding migrates from prefrontal cortex to dorsal striatum during learning of a self-timed response duration task
Time is essential to behavior, yet how we learn to time an interval is unclear since most behavioral timing tasks require weeks of training. Using a novel rapid training task, we captured the emergence of timing signals in rats. Machine learning and pharmacological tests revealed a double dissociation in temporal learning: the medial prefrontal cortex drives early learning, while the dorsal striatum engages as animals become proficient in the task. eLife
An open-source, ready-to-use and validated ripple detector plugin for the Open Ephys GUI
Sharp wave-ripples are key hippocampal events related to learning and memory that occur shortly in time (40 ms). Causal manipulations of these events are hindered by a lack of replicable detection systems and by a high false-positive rate for use during behavior. To tackle this, we developed an open-source, ready-to-use Open Ephys plugin with built-in movement filtering to reduce false detections. Simulations and closed-loop tests showed high accuracy, low false positives, and fast latency, providing a reliable platform for ripple-based neuroscience experiments. Journal of Neural Engineering
Long-duration hippocampal sharp wave ripples improve memory
Hippocampal sharp wave-ripples are essential for memory, yet how their properties support learning has been largely unclear. We found that ripple duration increases when rats learn a new spatial task. Artificially prolonging these naturally occurring events enhanced memory, whereas shortening them impaired it. In contrast, randomly triggered ripples carried minimal spatial content. These results show that extended, naturally initiated ripples play a critical role in memory. Science
More publications on my Google Scholar profile