Marius Pachitariu, Ph.D.
Mind-reading from orofacial behaviors in mice
The brain never stops. Our 100 billion neurons continuously fire action potentials, and most of that neural activity never leaves the brain to drive muscles and behavior. What does it mean? Are neurons merely maintaining homeostasis, are they “hallucinating” recent sensory experiences, or does the neural chatter play a bigger role to the brain’s function? To find out, we recorded populations of up to 25,000 neurons from the visual cortex of mice sitting in complete darkness. We found a highly-elaborate statistical structure of hundreds of dimensions of co-variability between neurons. A large fraction of this spontaneous activity could be predicted on a moment-by-moment basis from the spontaneous orofacial behaviors of the mice. We then used simultaneous recordings from eight Neuropixels probes totalling 3072 channels and found that the entire brain engages in these neural activity patterns. The activity continued uninterrupted during sensory stimulation and did not affect sensory coding: we could decode the orientation of a visual stimulus to an average 0.3 deg of precision, which is much more precise than mouse behavioral performance. Furthermore, the spontaneous neural patterns were completely orthogonal to the patterns evoked by sensory stimulation, arguing against the hallucination hypothesis. We conclude that a whole other layer of internal neural processing is overlaid on top of every brain area, minimally affecting its normal function.
Marius Pachitariu has a B.A. in mathematics from Princeton University, and a Ph.D. in computational neuroscience and machine learning from the Gatsby Unit at UCL. Following his doctoral studies, he joined the cortex lab at UCL where he developed experimental and computational techniques for large-scale neural recordings in mice, using calcium imaging and electrophysiology. He is now a group leader at HHMI’s Janelia Research Campus, where he continues investigating large-scale neural dynamics and developing new methods for data analysis. He is an advocate of open science and open source and shares his lab’s data and code freely with the community, always in advance of publication.