Scientists Develop Host-Aware Framework to Predict Gene Expression Across Contexts

For years, synthetic biologists have faced a stubborn problem: the genetic parts they use to build living systems don't behave the same way twice. A promoter that drives strong protein production in one bacterial cell might sputter in another. A ribosome binding site that works beautifully in one genetic context can fail mysteriously in the next. This inconsistency has made it nearly impossible to design gene circuits with real predictive power. You could measure how a part performed in isolation, but those measurements told you almost nothing about what would happen when you dropped that same part into a different host cell or a different physiological state.

A team of researchers has now developed a framework that tackles this fundamental challenge head-on. Rather than treating bioparts as isolated components with fixed properties, they built a system that accounts for the host cell itself—the living context in which those parts must function. The approach centers on a digital twin of Escherichia coli, a computational model that mirrors the bacterium's actual physiology and can be tuned based on measured growth rates. By combining this digital twin with carefully designed combinatorial libraries of plasmid-based expression constructs, the researchers were able to separate what a biopart inherently does from what the host cell is contributing to the outcome.

The work focused on three critical types of bioparts: plasmid origins of replication, promoters, and ribosome binding sites. Using structured combinatorial libraries—essentially systematic collections of genetic variants—the team identified parameters for each part that are mechanistically interpretable and, crucially, transferable across different conditions. They discovered that ribosome binding sites have what they call an intrinsic translation initiation capacity, a fundamental property that captures the dominant contribution of the RBS to translation. Everything else—the variation in expression that appears when you move from one physiological condition to another—emerges from the host's physiology and the local sequence context surrounding the part.

This distinction matters enormously. Once you know a part's intrinsic properties, you can predict how it will behave in new contexts by accounting for host-dependent effects. The parameterization the team developed proved accurate at predicting protein synthesis across a range of physiological conditions. It also supported incremental expansion of their libraries, meaning researchers can keep adding new variants and the model stays predictive. Perhaps most importantly, the framework revealed where modularity breaks down—where bioparts stop behaving as independent units and start interfering with one another in ways the simple models couldn't capture.

The research was supported by funding from Spain's Ministry of Science and Innovation, the Universitat Politècnica de València, Ecuador's higher education ministry, and the European Commission's NextGenerationEU program. The work represents a significant step toward what the researchers call host-aware design in synthetic biology, a more sophisticated approach that acknowledges the living cell as an active participant in determining gene expression rather than a passive container.

What this framework provides is a scalable foundation for building synthetic gene circuits with genuine predictive power. Instead of designing in the dark and hoping parts work together, synthetic biologists can now model how their designs will actually behave before they build them. The approach doesn't eliminate the need for experimental validation, but it transforms the design process from guesswork into something closer to engineering. As synthetic biology moves toward more complex applications—from biosensors to biofuels to therapeutic cells—this kind of predictive capability becomes not just convenient but essential.