Imagine having cells emit radio waves, enabling real-time readings of various biomarkers in living organisms. This would mean real time diagnoses, and much cleaner and more powerful screening.
We could engineer cell circuitry that triggers luciferin expression, subsequently reacting with oxygen and emitting waves with frequencies in radio range. We could design multiple such circuits in a single cell, each linked to a different luciferin. The emitted signal, consisting of radio waves at distinct frequencies, could be processed to extract multi-dimensional features. Radio-waves can probably be read from far away and get through several layers of cells too!
The hurdle to making this happen is the lack of natural luciferins that operate within the radio range. Although most natural bioluminescence falls within the visible spectrum, that may be more of an evolutionary quirk rather than a fundamental limitation — there was probably no advantage to emitting waves no organism could see!
One approach would be to slap ML on the problem, and try to find a direction in the latent space of a large protein language model that correlates with wavelength. We have ~11 different luciferins, so the dataset is incredibly sparse. Unlikely that a zero-shot approach works. You could just mutate the existing luciferins to potentially get a larger dataset, but the percentage of candidates that emit something you end up creating is probably really tiny too.
A more likely to work approach would be to evolve these proteins. You probably want some form of continuous evolution (Phage assisted continuous evolution can do 100 cycles in ~2 weeks!) — where you start with some existing luciferin and apply pressure to increase the wavelength. Seems doable to make the circuit for continuous evolution: chromophores are already used in the natural world to sense light of different wavelengths. So you just have to find a way to link the ability to produce high wavelength emissions with a phage’s fitness.
Once you have a bunch of luciferins you’d like to use, you can probably design a common luciferase to catalyze all of their reactions with oxygen! The Baker Lab recently showed an example of this using RFDiffusion.