Closed-Loop Discovery for Synergistic Antibacterial Combinations

Research Question

How can an experimental system choose which antibacterial combinations, concentrations, or conditions to test next — under assay cost, feasibility, and safety constraints — so that each experiment maximises information about synergistic interactions rather than merely screening exhaustively?

Method

This project develops uncertainty-aware active experiment selection for antibacterial combination discovery. The approach connects three methodological pillars from the Methods page:

• Adaptive decision-making — prioritising experimental conditions that are expected to be most informative given current evidence and uncertainty, under safety and cost constraints.
• Uncertainty-aware inference — using calibrated uncertainty to distinguish well-established findings from exploratory results, avoiding premature conclusions from sparse data.
• Closed-loop experimentation — connecting automated selection, execution, and inference into a feedback loop where expert review gates actions that require safety judgment.

Platform scope

The platform targets synergistic antibacterial combination discovery, with potential for hydrogel-based topical delivery. A related collaboration applies instruction-tuned language models to mine biomedical interaction entities from multi-source evidence.

Personal Contribution

Frederik Banis contributes the uncertainty-aware decision and inference methodology — framing experiment selection as a sequential decision problem under uncertainty, connecting it to the broader adaptive sensing and control framework developed in prior energy-systems work.

Validation and Key Results

• Manuscripts describing the platform and related work are under review. They are not counted as published outputs.
• Specific quantitative results are pending peer review and are not reported here to avoid presenting proposed outcomes as completed work.
• Detailed results and the full platform description will be available upon publication.

Outputs

• Manuscripts (under review): Details will be disclosed upon publication.
• Software: None released at this time.

Collaborators

• Collaborating research team at Hangzhou Institute of Technology, Xidian University.
• Further collaborator details will be provided upon publication of the associated manuscripts.

Status and Next Steps

Status: Ongoing. Manuscripts are under review; none are counted as published. Future closed-loop experimentation directions remain proposed or ongoing and are not presented as completed achievements.

Next steps:

• Advance the manuscripts through peer review.
• Extend the active experiment selection framework to additional biomedical collaboration domains.
• Integrate calibrated uncertainty estimates into the selection loop to distinguish established from exploratory findings.

Connection to prior work: The closed-loop experimentation cycle — measure, infer, decide, act — with an expert review gate is the same architecture validated in hardware-in-the-loop microgrid control (published). The shift is from energy hardware to biomedical hardware, but the closed-loop validation methodology is shared.

See the Methods page for the closed-loop experimentation pillar, the Research page for the research program overview, and the Applications page for the full application map.