AI guided biotronic inter-kingdom interoperability: a universal adaptor for life.
Abstract
Publicly Releasable Abstract All cell types, in all species, communicate using evolutionarily ancient bioelectric signaling mechanisms. However, much remains to be learned about these mechanisms, which charged species participate in which ways in computational and control networks, and how shared or divergent these mechanisms are across different taxa. To investigate these questions we propose to develop AI methods that can learn to rearrange living components to facilitate bioelectric communication across a single organism (Aim 1), then a conspecific colony (Aim 2), and finally between organisms from different kingdoms (Aim 3). The AI method will be composed of three components. The first will be a pair of computational models of the two model organisms selected for study: (1) ÒxenobotsÓ, which are mm-sized self-motile rearrangements of early frog embryo; and (2) bacterial biofilms composed of different admixtures of wild type and mutant strains. The second AI component, the hypothesis generator, will search the space of all hypothetical bioelectric signaling mechanisms that might be at work in either species. This method will find a set of these hypotheses that, when added to the baseline computational model, allow the model to reproduce all of the observations of the bioelectric behavior of the model organism so far. The third AI component, the experiment generator, will search over the space of experiments: new xenobots or bacterial biofilms that could be constructed with the performersÕ existing experimental facilities. The experiment generator will search for interventions that, when supplied to the current set of computational model/hypothesis pairs, causes them to maximally disagree in their predictions of the resulting transmission (or lack thereof) of bioelectric signal propagation within or between the two model organisms, if those organisms were constructed. The most disagreement-inducing of these AI-designed organisms will be built and their ability to transmit bioelectric signals internally or externally will be recorded. Armed with this expanded set of observations, another round of hypothesis and experiment generation will be conducted, leading eventually to accurate and human-interpretable models of bioelectric behavior in these two organisms, ways to exploit that behavior, and AI-designed organisms from different kingdoms capable of bioelectrically signaling to each other. Future work could extend this method to enable diverse species to not just to signal to one another, but communicate behaviorally-useful information between themselves. If successful, this will result in species-independent methods for creating bioelectronic interoperability between living systems with diverse functions, thus catapulting synthetic biology forward much like electronic interoperability did for information technologies. Viewed differently, our AI method is the beginning of a Robot Scientist roadmap to uncover deep design principles of biological computation and information processing. This will be enabled by interrogation of unconventional (hybrid) biological systems, which allows inference of general rules not constrained by frozen accidents of evolutionÕs prior history.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 19, 2023
- Source ID
- W911NF2310100
Entities
People
- Josh Bongard
Organizations
- Army Contracting Command
- United States Army
- University of Vermont