#

Defining the structure and function of brain regions is a hard practice, which is usually conducted manually. Spatial protein patterns present a data-driven approach to exploring the organization of the Drosophila fly brain. Several previous studies showed promising results using limited datasets. Here we introduce an algorithm for dimensionality reduction and a revised version of a clustering procedure, both using the power of graphic processor unit technology for high-speed performance. Our segmentations, based now on more protein signals, were compared to a classical reference template of the fly brain, at two levels of granularity. Results were found to be more stable, and they suggest promising relations between the molecular and structural knowledge of the Drosophila melanogaster brain.