Brain hypoxia affects hundreds of thousands of people annually, leading to perinatal brain injury, complications of ischemic strokes, vascular dementia, and neurodegeneration. Despite this, there are no FDA approved drugs specifically targeting brain hypoxia. Regeneration of capillaries, or therapeutic angiogenesis, is a strategy to treat patients with brain hypoxia. One way to accomplish this is by enhancing or mimicking the body’s own angiogenic response. A challenge lies in understanding this response with the mechanistic-detail needed to develop effective therapeutic strategies. Methods: Two main growth factors that regulate angiogenic sprouting in the brain are vascular endothelial growth factor (VEGF) and brain-derived growth factor (BDNF). We present an integrated computational-experimental strategy to characterize human endothelial cell-cell interactions as a function of these proteins. We introduce a 3D “rules-as-agents” model of cell movement that allows rapid testing and comparison of multiple hypotheses to in vitro angiogenesis experiments. Endothelial cells are represented as machines that transition between finite behavior states, and their properties are explored by a genetic search algorithm. We rank and quantify differences between competing hypotheses about cell behavior during the formation of unique capillary phenotypes. We validated the ‘state-machine’ paradigm through in vitro assays. In vitro assays included 3D angiogenic spheroid assays and the 2D classification of primary human umbilical vein endothelial cell (HUVEC) states as a function of varying concentrations and combinations of BDNF and VEGF (control; 25 ng/ml VEGF; 50 ng/ml VEGF; 25 ng/ml VEGF, 25 ng/ml BDNF; 25 ng/ml VEGF, 50 ng/ml BDNF; 50 ng/ml VEGF, 100 ng/ml BDNF; 50 ng/ml BDNF; 100 ng/ml BDNF), at 6, 12, 24 and 48 hrs of stimulation. Hierarchical and k-means cluster analysis coupled to molecular imaging identified similarities between stimulated HUVEC based on 7 main categories including intracellular cytoskeletal regulation, actin fiber alignment, and network connectivity. Results: Our best adaptive state machine models of endothelial cell behavior as a function of BDNF and VEGF predicted novel in vitro results within a standard deviation of all measurements (22% of models). Cluster analysis of >10,000 endothelial cells revealed a set of dominant morphological phenotypes and cell-cell connectivity. These results show that VEGF and BNDF can reproducibly induce unique frequencies of behavior; morphological angiogenic states; and cell-cell connectivity in 2D and 3D. Ongoing work involves delineating the effect of these identified endothelial cell phenotypes and networks on neural function. This work offers the ability to understand – and ultimately control – human cell behavior at the microvasculature level. Applications of these technologies include guiding regenerative therapies targeting the neurovasculature.