Alexandra VandeLoo
BioE Ph.D. Proposal Presentation
Time and Date: 9:00 am, October 30th, 2025
Location: MRDC 3515


   Advisor: Dr. Craig Forest (Georgia Institute of Technology)

   Committee Members: 
   Dr. Scott Danielsen (Georgia Institute of Technology)
   Dr. Levi Wood (Georgia Institute of Technology)
   Dr. Julie Champion(Emory University)
   Dr. Todd Fernandez (Georgia Institute of Technology)

Integrating AI-Driven Cell Quantification and Electrophysiology for Accelerated Protein Engineering
Optogenetics has revolutionized neuroscience by allowing the precise control of neuronal activity using light-sensitive ion channels, yet the experimental processes underlying its application remain difficult, variable, and resource-intensive. The cultivation of healthy neurons, acquisition of electrophysiological skills, and engineering of improved optogenetic tools each present steep technical and conceptual challenges that slow research progress. This dissertation integrates computational image analysis, educational simulation, and protein engineering to improve standardization, training, and discovery, optimizing the pipeline of cellular neurophysiology.
To improve cell health, we developed a quantitative image-analysis platform for automated segmentation and morphological assessment of cultured neurons. This software enables objective quantification of cellular health, density, and morphology, improving the reproducibility of culture preparation and providing a rigorous foundation for downstream transfection and electrophysiological studies. We then designed an interactive patch-clamp training simulation that replicates the visual and mechanical cues of real experiments. Patch clamping, a Nobel Prize–winning technique essential to measuring neuronal ion channel function, typically requires months of supervised practice. In controlled evaluations, the simulation improved both user performance and confidence, offering a scalable, low-cost solution for electrophysiology training. Building on these validated experimental and educational foundations, we next applied protein language models to guide mutational design of channelrhodopsins—the light-gated ion channels that enable optogenetic control. Mutations predicted by the models were synthesized, expressed in HEK293FT cells, and functionally characterized using whole-cell patch clamp under light stimulation. Several model-guided variants exhibited altered photocurrent amplitudes and kinetics, demonstrating the potential of machine learning to reveal non-intuitive sequence–function relationships and accelerate rational protein engineering.
Together, this work presents a unified, technology-driven framework that enhances experimental precision, training efficiency, and protein discovery in neuroscience. By combining automated cell quantification, interactive electrophysiology education, and machine learning–guided protein design, this dissertation advances the pipeline and workflow for single-cell electrophysiology and protein engineering.