Abstract
Background and Purpose
Neurotoxicity is a major factor in drug development failures, accounting for 12% of drug withdrawals between 1960-1999. Complex in vitro models (CIVM) derived from human tissue offer improved clinical translation and scalability for early drug screening. This study aimed to stabilize and improve the reproducibility of neurotoxicity predictions using a cortical organoid platform with automated analysis.
Methods
- 3D cortical organoids were derived from healthy donor iPSCs, differentiated into neuronal progenitor cells (NPCs), and seeded in 384-well plates.
- After 10 weeks of differentiation, spontaneous network activity was recorded using a calcium flux assay and a high-throughput kinetic plate reader (FLIPR).
- Automated analysis pipelines were developed to refine neurotoxicity predictions by improving peak detection, waveform feature engineering, and automated potency calculations.
Results
- The model achieved high specificity (≥90%) and good sensitivity (>50%), replicating prior results.
- Predictions remained stable across multiple experiments conducted in different labs over several years using distinct cell banks.
- Automation improved reproducibility by removing user bias and enhancing waveform feature analysis.
- Key waveform features like peak shape, frequency, and decay time were leveraged to improve detection of neurotoxic responses.
Conclusion
This enhanced cortical organoid platform offers a robust, scalable method for early-stage neurotoxicity screening. By improving model stability and reproducibility, it can reduce costly clinical failures and improve drug candidate selection in preclinical testing.
