We investigate the organizational principles of neuronal circuits. Our goal is to understand the mechanisms that shape neuronal circuits and the ways in which their structure supports complex computations.

To achieve this, we develop computational approaches for reconstructing, annotating, and analyzing connectomics datasets at synaptic resolution.

Research Areas

Connectome-enabled analysis of neuronal circuits

The advent of synaptic wiring diagrams is revolutionizing how neuronal circuits are analyzed. In the fruit fly, where a whole-brain connectome is now available, almost all analyses use the connectome in some form. The same revolution is coming to the mammalian circuits for which partial connectomes spanning up to 1 mm³ have just been created. We devise new analyses and approaches to interpret these connectomes, extract the organizational principles of mammalian circuits, and model how their structure supports complex computations.

Scaling to connectomes of whole mammalian brains

As imaging technologies have advanced to acquire larger samples of brain tissue at synapse resolution, connectomics is increasingly bottlenecked by our ability to reconstruct and annotate the neuronal circuits within them. Current approaches are sufficient to reconstruct fly-scale connectomes with 10s of person-years of manual labor, but are still too costly even to reconstruct the largest connectomics datasets to date. To scale connectomics to whole mouse brains, we work on a new generation of machine learning approaches for the reconstruction of neurons and annotations of all their features.

Towards multi-modal circuit analyses: Linking connectomes to other modalities

Comprehensive knowledge about the cell types that constitute the molecular and anatomical architecture of the brain is crucial to understanding how brain functions emerge from its circuits. Transcriptomic information enables us to link cells to specific molecules and genetic tools. We develop machine learning approaches that bridge between connectomics datasets and cellular datasets of other domains, such as patch-seq, to create multi-modal connectomics datasets.

Computational and team science approaches for connectome reconstruction

Reconstruction, annotation, and analysis of connectomic datasets is a complicated interplay between computational models and communities of human experts. We develop computational approaches to enable these collaborative communities and make connectomics datasets broadly available and analyzable. For instance, together with our collaborators at the Allen Institute for Brain Science and Zetta AI, we develop and maintain the Connectome Annotation Versioning Engine (CAVE), a software infrastructure that hosts many connectomics datasets across the connectomics field. CAVE provides a backbone for connectomics analyses by enabling scalable solutions for proofreading and flexible annotation support for fast analysis queries at arbitrary time points.