Project Summary Estimates of the total length of axonal "wiring" in the human brain are on the order of hundreds of thousands of kilometers. Understanding the fundamental principles underlying the connectivity between cells is a daunt- ing task, but it has become increasingly clear that there are canonical connectivity patterns across the layers of the mammalian cortex. Many of these pairwise connectivity rules between cells have been discovered using multi-patching in slices, but examining higher-order connectivity motifs (for example, triangular motifs) is difficult in the slice preparation. Furthermore, a central explanatory goal of neuroscience is to relate functional proper- ties of neurons to the underlying connectivity between them. Achieving this goal requires overcoming significant technical challenges, but a few heroic studies have managed to identify such functional/structural principles such as enhanced "like-to-like" connectivity in visual cortex cells that prefer similarly oriented stimuli. Over the past five years, our team has participated in a "moon-shot" project as part of the IARPA and BRAIN Initiative-funded MICrONS project to collect functional and synaptic-scale anatomical data from a millimeter cube of mouse visual cortex. Functional in vivo calcium imaging of this volume was performed at Baylor College of Medicine in Hous- ton, then the mouse was shipped to Seattle where the same volume was extracted, prepared, sliced at 40nm thickness, and imaged on an array of advanced electron microscopes. Finally, the approximately two petabyte image stack was finely-aligned and segmented by Sebastian Seung's group at Princeton. Achieving this ambi- tious goal took almost the entire five years of the MICrONS program which ended in July 2021. This data set has now beeen shared with the entire neuroscience community and has huge untapped potential for scientific discovery. In Aim 1 we will use graph theoretical methods and focus our analysis to identify local higher-order circuit motifs across layers and large-scale modules between excitatory neurons across cortical layers focusing in mouse V1. We will test the hypothesis that groups of excitatory neurons form tightly-connected modules with sparse, reciprocal connections to other modules. In Aim 2 we will focus on relating structure to function. At the local circuit level we will characterize the relationships between stimulus selectivity and connectivity within and across cortical layers in V1. We will test the hypothesis that connected groups of neurons (i.e. structural modules) form computational modules to represent similar stimulus preferences (such as textures). For these analyses we will leverage validated deep learning predictive models that provide a flexible, systematic method to characterize even non-classical, non-linear feature selectivities of neurons and find the neuron's most-exciting inputs.