This dissertation investigates the physical constraints and organizational principles that shape real-world networks, with a focus on neural systems. Physical networks such as neurons or vasculature are spatially embedded and subject to volume exclusion, meaning their edges cannot physically cross through one another. As a result, understanding these systems requires more than just adjacency matrices, because spatial layout plays a central role. The first part of the dissertation examines optimization principles that govern physical networks. While total edge length minimization is a common framework to understand organization in biological systems, it does not fully explain key features of biological branching. In this talk, I present collaborative work showing that a higher-order principle known as manifold minimization, which treats edges as physical tubes and seeks to minimize surface area, provides a better explanation for observed patterns. This framework accounts for branching angles and trifurcations across a range of biological systems, and even offers explanations for system specific features like synaptic connectivity. The second part focuses on neuronal networks and their structural properties across species. Analyzing eight brain datasets from five organisms, I find that both degree and synapse count distributions follow lognormal patterns. To investigate their origin, I reduce 3D neuron meshes to skeletons and study the growth of neurites. I propose that neurite length arises from a multiplicative branching process, offering both a generative explanation for connectivity and synaptic organization, and a mechanism for lognormal networks. Further, using this newfound understanding of neural connectivity, we develop a null model for neuron connectivity in the brain.These findings support more biologically grounded models of brain structure and function.