Applied Pharmaceutical Toxicology (APT) 2025 Conference
May 13–15, 2025 | San Francisco, CA
Preclinical Neurotoxicity Predictions using a
High-Throughput Neural Spheroid Model
Speaker Name
David A. Gallegos, PhD, Senior Scientist, Takeda
Abstract
The drug development process is fraught with failure due to either safety issues or poor efficacy. When considering safety profile, neurotoxicity is the leading cause of clinical failure. Furthermore, 12% of drugs withdrawn between 1960-1999 were caused by CNS-related adverse events. The use of advanced in vitro models (AIVM) derived from human tissue has dramatically expanded in recent years, promising to provide the necessary biological complexity to improve clinical translation and scale to enable adoption early in drug development pipelines. We have worked with AxoSim on the development of and modeling around a cortical brain organoid model that exhibits robust spontaneous calcium waveform activity that is compatible with HTS methodology and provides a clinically relevant endpoint for phenotypic profiling. In 2022, this organoid platform was used to develop a predictive clinical neurotoxicity model that showed remarkable specificity (>90%) and good sensitivity (>50%), making it an ideal pre-screening method prior to standard 2-species animal testing. Here, we tested the stability and reproducibility of these predictions over time and used these replicate experiments to refine and automate neurotoxicity score predictions
Improving the stability and reproducibility of clinical neurotoxicity predictions from a high-throughput compatible neural organoid platform
Overview
This study, presented by researchers from AxoSim Inc. and Takeda Pharmaceuticals, evaluates the robustness and reproducibility of a 3D human cortical brain organoid platform for predicting clinical neurotoxicity. Originally developed in 2022, the model has demonstrated high specificity and moderate sensitivity in identifying neurotoxic compounds, offering an early preclinical screening alternative to animal models.
Key Highlights:
1. Purpose & Background:
- Neurotoxicity is a major contributor to drug failure during development.
- The study builds on previous work using human iPSC-derived cortical organoids that replicate spontaneous neural activity.
- The goal was to improve prediction stability via automated analysis pipelines and confirm reproducibility across labs and cell sources.
2. Methods:
- Organoids were generated from iPSCs and matured over 10 weeks.
- 84 compounds (both neurotoxic and safe) were tested using a calcium flux assay.
- Automated tools quantified six key waveform features (e.g., frequency, rise time, spacing).
- Data normalization and machine learning (logistic regression) were used to predict toxicity.
3. Results:
- The system achieved ≥90% specificity and ~53% sensitivity, consistent with earlier findings.
- Predictions remained stable across multiple studies, sites, and iPSC sources.
- Toxic and safe compounds were clearly distinguished based on waveform modulation and margin of exposure.
4. Automation & Analysis Improvements:
- New automated tools improved consistency and removed user bias from peak detection.
- Custom software calculated neurotoxicity prediction scores based on real-time waveform changes.
Conclusion
- The neural organoid platform reliably predicts clinical neurotoxicity and is suitable for high-throughput screening.
- It can help pharmaceutical developers de-risk candidates earlier, reducing reliance on animal models and minimizing costly late-stage failures.