Automated assembly and digital twins of biological neurosphere networks for computation and disease modeling

Principal investigator: Victoria Webster-Wood

Co-PI: Jessica Zhang

University: Carnegie Mellon University

Industry partner: HexSpline3D, LLC

Neurospheres and brain organoids can replicate key brain-network functions in vitro, but culture sensitivity and stochastic maturation make interconnected networks difficult to manufacture reproducibly. This one-year project will integrate automated assembly with data-driven digital twins to build and predict multi-neurosphere circuits for biocomputing, disease modeling, and drug screening. Task 1 will generate size-controlled mouse cortical neurospheres (under 300–350 micrometers) and fabricate Plant Design Management System modular masks with asymmetric microchannels to bias unidirectional axon growth, then automate neurosphere placement via a 3D printer and G-code workflow and quantify connectivity using differential viral labeling with time-lapse and confocal imaging. Task 2 will train a MetaFormer model with spatiotemporal attention on these longitudinal datasets to forecast network formation and enable in silico screening of PDMS geometries prior to fabrication. The outcome is a higher-throughput, lower-cost pipeline for scalable neurosphere-network manufacturing.