Digital Worm (BAAIWorm)

BAAIWorm is a pioneering data-driven model that integrates the neural dynamics, body and environment of Caenorhabditis elegans into a closed‑loop system. By calibrating the coupled neural and biomechanical models against experimental data, the system reproduces characteristic locomotion and chemotaxis, providing a platform for testing links between neural structure and behavior.

The project was developed by the Life Simulation Research Centre at the Beijing Academy of Artificial Intelligence under the leadership of Dr. Lei Ma. It has been highlighted as a milestone in digital life research and appeared on the cover of Nature Computational Science in 2024.

Explore the project’s media coverage, including the CCTV News Live feature on BAAIWorm, the follow-up coverage on the Multimodal Trans-scale Biomedical Imaging Facility, and the original Nature Computational Science paper.

Model construction

BAAIWorm is an open‑source, modular system composed of two data‑driven submodels: a biophysically detailed neural network and a lifelike body–environment model. Together they provide a versatile platform for investigating how neural circuits control embodied behaviour.

• Neural network. The published closed-loop experiments use 136 biophysically detailed multi-compartment neuron models connected by chemical synapses and gap junctions. The full hermaphrodite nervous system has 302 neurons; modeling all of them is an extension beyond the published closed-loop model.
• Body & environment. The soft-body model uses 3,341 tetrahedral elements and 96 muscle units and interacts with a three‑dimensional fluid environment that exerts thrust and drag. Continuous sensory inputs, such as a food concentration gradient, modify neural computations and drive muscle contractions.
• Closed‑loop feedback. Sensory signals from the environment feed back into the neural network, which in turn drives muscle activation to generate movement. This closed loop aligns the worm’s behaviour with environmental cues.

Research findings

The team optimized connection weights and polarities against experimentally measured neural population dynamics, then coupled the neural model to body and environment dynamics. The resulting closed-loop model reproduces characteristic zigzag locomotion toward an attractor, linking neural activity to observable behavior.

To probe causal relationships, synthetic perturbation experiments altered synaptic and gap‑junction properties at specific sites or across the entire network. These manipulations changed neural dynamics and disrupted coordinated movement, demonstrating how the structure of the connectome governs behaviour.

Research recognition & impact

BAAIWorm appeared on the December 2024 cover of Nature Computational Science and was the subject of the journal’s Research Briefing, “A simulated C. elegans with biophysically detailed neurons and muscle dynamics”.

The 2026 review The embodied brain (preprint) by EPFL researchers Sibo Wang-Chen and Pavan Ramdya lists BAAIWorm among its papers of outstanding interest. It discusses the model’s connectome-constrained neural circuitry, anatomically grounded muscles and closed-loop behavior.

National coverage includes Xinhua’s March 2026 interview with Lei Ma and CCTV’s feature on the Digital Worm project.

Enduring impact & future directions

As an early result of the Digital Life initiative, BAAIWorm demonstrates that faithfully reproducing the interaction of neural dynamics, biomechanics and environment is both feasible and informative. Its open‑source and modular design allows researchers to extend and replace components, making the platform transferrable to other species and fostering collaboration. By providing a reusable framework for closed‑loop simulation, BAAIWorm paves the way toward digital organs, digital animals and ultimately a broader digital life programme.

Future work will incorporate additional connectome and behavioural data to improve biological realism and expand to more complex sensory modalities.