An AI-Cyborg System for Adaptive Intelligent Modulation of Organoid Maturation

Ren Liu, Zhaolin Ren, Xinhe Zhang, Qiang Li, Wenbo Wang, Zuwan Lin, Richard Lee, Jie Ding, Na Li, Jia LiubioRxiv 2024 · DOI

TL;DR

Researchers have developed an "AI-cyborg" system that integrates flexible bioelectronics with artificial intelligence to precisely control and accelerate the maturation of human cardiac organoids. This closed-loop system learns optimal stimulation patterns in real-time, leading to significantly improved functional and structural development compared to traditional methods.

What Problem Does This Paper Address?

Developing functional, mature organoids for research and therapeutic applications is a major goal in regenerative medicine. However, in vitro organoid development often lags behind in vivo maturation, leading to immature cells with potential risks like tumor formation or reduced drug efficacy. While bioelectronic interfaces can monitor and stimulate organoids, effectively using the collected data to dynamically guide maturation has been a significant challenge. Conventional stimulation methods are often static and not adaptive to the evolving state of the organoid, limiting their effectiveness.

Methods

The core of this work is the BIO-AIM (Bioelectronics Interface with Optimized Adaptive Intelligent Modulation) system. This system combines:

  • Tissue-like, Flexible Bioelectronics: Ultra-flexible mesh electronics are integrated directly into 3D cardiac organoids. These electronics act as both sensors (recording electrical activity) and actuators (delivering electrical stimulation).
  • Cyber-Learning Framework: An AI agent, specifically using a Bayesian Optimization (BO)-inspired Reinforcement Learning (RL) approach, drives the system.
  • State Representation: The organoid's developmental stage is represented by "pseudotime," inferred from its electrical activity using a dimensionality reduction technique.
  • Action Space: The AI agent controls stimulation parameters like frequency and amplitude.
  • Reward Function: The AI is rewarded for actions that advance the organoid's pseudotime, effectively accelerating maturation.
  • Learning Algorithm: A Gaussian Process (GP) model is used to predict the effect of stimulation on pseudotime, and BO is employed to efficiently explore the parameter space and find optimal stimulation policies, especially crucial given limited experimental data.
  • Closed-Loop Feedback: The system continuously monitors the organoid's electrical activity, uses this data to update the AI's understanding of the organoid's state, and then adjusts stimulation parameters in real-time.

Datasets

Not explicitly discussed as training datasets in the traditional machine learning sense. The system relies on real-time experimental data collected from human pluripotent stem cell-derived cardiac organoids (hPSC-CMs). This data is used to continuously update the AI's model and control policy.

Key Insights and Ideas

  • AI as a "Nerve Innervation" Mimic: The AI-cyborg system effectively mimics the role of nerve innervation in natural organ development, providing continuous, adaptive sensing and stimulation that is typically absent in in vitro systems.
  • Pseudotime as a Maturation Metric: Using pseudotime derived from electrical activity provides a powerful, continuous, and data-driven metric for tracking organoid maturation, enabling sophisticated AI control.
  • Bayesian Optimization for Sample Efficiency: The use of BO is critical for learning effective stimulation policies with limited biological experimental data, a common bottleneck in organoid research.
  • Adaptive Stimulation Outperforms Static: The AI-driven adaptive stimulation strategy demonstrably leads to superior organoid maturation compared to both fixed stimulation protocols and no stimulation.

Key Results

The BIO-AIM system successfully accelerated the maturation of hPSC-CM organoids, validated by:

  • Enhanced Electrical Activity: Increased extracellular spike waveform amplitude and maximum dV/dt.
  • Faster Conduction Velocity: Mean conduction velocity reached 102.6 ± 2.3 mm/s in the AI-driven group, significantly higher than the fixed stimulation group (60.6 ± 1.5 mm/s).
  • Improved Sarcomere Organization: Immunostaining revealed well-aligned sarcomere structures in AI-stimulated organoids, outperforming no-stimulation and fixed-stimulation groups.
  • More Uniform Maturation: AI-driven stimulation reduced batch-to-batch variation in electrical function.

Limitations and Future Work

The authors acknowledge several areas for future development:

  • Explaining the AI Policy: Further studies are needed to understand why the AI-learned stimulation policies accelerate maturation and to determine the optimal parameters.
  • Mechanism Elucidation: Detailed analysis of gene/protein expression, cell connectivity, and morphology is required to understand the underlying mechanisms of AI-driven maturation acceleration.
  • Hardware Enhancements: Scaling up sensor/actuator density and integrating application-specific integrated circuits (ASICs) could improve resolution and accuracy.
  • Software Enhancements:
    • As more data becomes available, more advanced representation-learning techniques could be employed.
  • For more complex tissues (e.g., brain organoids), a network policy that handles different cell types and locations will be necessary, potentially requiring multi-agent RL.
  • Considering long-term effects of actions might require adapting techniques for higher sample complexity.
  • Incorporating multimodal sensors and stimulators (e.g., chemical, mechanical) would enable a more comprehensive AI agent.

Summary updated

Abstract

Abstract Recent advancements in flexible bioelectronics have enabled continuous, long-term stable interrogation and intervention of biological systems. However, effectively utilizing the interrogated data to modulate biological systems to achieve specific biomedical and biological goals remains a challenge. In this study, we introduce an AI-driven bioelectronics system that integrates tissue-like, flexible bioelectronics with cyber learning algorithms to create a long-term, real-time bidirectional b ioelectronic interface with o ptimized a daptive intelligent m odulation (BIO-AIM). When integrated with biological systems as an AI-cyborg system, BIO-AIM continuously adapts and optimizes stimulation parameters based on stable cell state mapping, allowing for real-time, closed-loop feedback through tissue-embedded flexible electrode arrays. Applied to human pluripotent stem cell-derived cardiac organoids, BIO-AIM identifies optimized stimulation conditions that accelerate functional maturation. The effectiveness of this approach is validated through enhanced extracellular spike waveforms, increased conduction velocity, and improved sarcomere organization, outperforming both fixed and no stimulation conditions.
BibTeX
@misc{liu2024ai,
  author = {Ren Liu and Zhaolin Ren and Xinhe Zhang and Qiang Li and Wenbo Wang and Zuwan Lin and Richard Lee and Jie Ding and Na Li and Jia Liu},
  title = {{An AI-Cyborg System for Adaptive Intelligent Modulation of Organoid Maturation}},
  year = {2024},
  doi = {10.1101/2024.12.07.627355},
  howpublished = {bioRxiv},
}
Figure