His target is the crops' photosynthetic machinery: first simplifying it as well as broadening its range, allowing it to make use of light beyond the red and blue that are naturally preferred. On longer timescales, he and his colleagues plan to redesign the way the captured energy is employed, using it to generate hydrocarbons rather than sugar.
Tweaking proteins to do human bidding is nothing new. Enzymes and antibodies, for example, have long endured such outrages. But that is not what Dr Ennist is up to. Rather than modifying existing proteins, his versions are being designed from scratch, using artificial-intelligence (AI) models, to be optimised for the task at hand. To start with, they would be inserted into a suitable organism, such as a plant or bacterium, to do their thing there. But eventually they could, he hopes, operate independently and thus form the basis of a new type of solar cell -- one that turns out petrol rather than electricity.
With this and other projects, ranging from artificial noses to covid-19 vaccines, the IPD, run by David Baker, joint winner of last year's Nobel prize for chemistry, is taking the much-hyped but under-delivering field of nanotechnology back to its roots. The future it once heralded of useful molecule-size factories has dwindled over the decades into a marketing gimmick for sunscreen ingredients and tennis-racket frames. Now, though, the original promise is back with a vengeance.
The new nanotech relies on three things. One is an ability to work out how a protein's structure affects its function (Dr Ennist is hunting ones capable of holding together the pairs of chlorophyll molecules that are the nub of photosynthesis in ways well-suited to capture light and transfer its energy to electrons). A second is to devise chains of amino acids (the building blocks of proteins) that would be expected to fold into the desired structure. And the third is to check computationally, before making them for real, that chains thus devised will indeed assume the target shape.
For the first of these tasks Dr Baker and his colleagues use RFdiffusion, an AI model they have developed to predict a protein's function from its structure. It does this in a similar manner to image-generating diffusion models, but with a training database of more than 200,000 natural proteins rather than photos and artwork.
For the second, their tool is ProteinMPNN, also trained in-house, which draws on databases of how amino acids interact with each other in protein chains and with other molecules that those chains encounter. And for the third they employ RoseTTAFold, a machine-learning model similar to software originally written by Dr Baker in the mid-1990s. So influential was this precursor that it inspired the creation of AlphaFold, a protein-folding AI model now backed by Alphabet's billions, and whose creators carried off the other half of the 2024 chemistry Nobel.
Once a design has been through this virtual mill, scientists can conjure it into existence by synthesising appropriate DNA and putting that into a bacterium or yeast. It can then be tested to see if it is truly up to the job.
Besides redesigning photosynthesis, groups at the IPD are working on a mind-bending array of other projects. These include circular protein fibres that could be linked up like mail armour to make novel fabrics; hybrid organic-inorganic materials (think snazzy versions of bone and mother-of-pearl); enzymes to digest hard-to-dispose-of plastics such as PET, thereby turning them into useful chemicals; and chip-based sensors that run molecules through protein pores to determine what they are. Technology of this kind already exists for DNA and its cousin RNA, but Dr Baker believes it can be applied to a far wider range of substances, creating devices that are, in essence, artificial noses. And these are just the non-medical applications.
All very well
In the field of health care, the opportunities are vast. The institute's covid vaccine, SKYCovione, for example, works by displaying synthetic copies of parts of the SARS-CoV-2 spike protein in a way that attracts the immune system's attention. IPD researchers have also created proteins they hope will transform the treatment of snake bites. These lock onto and neutralise venom molecules in the blood in the way that the antibodies now employed for that task do, but are smaller and easier to make.
Dr Baker and his colleagues have plans to attack Alzheimer's disease using a similar approach -- making proteins that bind to the molecular precursors of the neuronal plaques and tangles found in the brains of those afflicted. And they hope to improve the field of gene editing with custom-targeted nucleases, the "Cas" part of the CRISPR-Cas complexes which are gene-editing's molecular scissors. These would be designed to bind to particular DNA sequences, increasing the range of DNA that can be edited and reducing the risk of off-target edits.
Where Dr Baker has led, others are following. Alphabet has two ongoing protein-design projects spearheaded by Sir Demis Hassabis, one of AlphaFold's Nobel-winning inventors. One, Isomorphic Labs, in London, is a spin-out that has contracts with the pharmaceutical firms Eli Lilly and Novartis to test candidate drug molecules' interactions with target proteins. The other is AlphaProteo, a system developed by Google DeepMind to design proteins to bind to specified targets.
Others are taking a slightly different tack. Profluent, in Emeryville, California and EvolutionaryScale, in New York, are building protein-design AI models that resemble not image-generating software, but large language models (LLMs) of the sort that power the world's chatbots. These firms' models treat the amino-acid sequences in protein chains like the words in a piece of text -- analysing relationships found in zillions of exemplars to design novel useful structures.
According to Ali Madani, Profluent's chief executive, the firm is particularly focused on creating new CRISPR-Cas gene-editing tools. Here, its USP is a curated database of around 5m CRISPR-Cas protein complexes on which its AI model has been trained in order to design new versions.
EvolutionaryScale is pushing the LLM approach still further. Its version, ESM3, takes into account a protein's structure and function as well as its amino-acid sequence. And its training database is huge. Alex Rives, the firm's chief scientist, says it contains 2.8bn entries. He also talks of going beyond working with individual proteins and creating a first approximation to a virtual cell, within which these proteins interact with one another.
In EvolutionaryScale's case, the model itself is the product, to be licensed to firms that plan to make protein-based drugs and materials. But many of its peers are pursuing innovation themselves. The consequences of this new approach to nanotech are as yet only dimly discernible. Redesigning photosynthesis, for example, would surely have consequences far beyond biofuels, particularly if the new approach could be made to work in existing plants. That, with due caveats for safety and customer acceptance, could boost crop yields. There is also huge scope for improvements in the yields of chemical processes: many enzymes are more efficient than conventional catalysts. And, as with any technology, less obvious breakthroughs may be possible, too.
One that excites Dr Baker is the idea of protein equivalents of the logic gates in silicon chips. These might be used to control gene expression in cells. In the longer term, he thinks, such gates could more easily be stacked in 3D arrays than their silicon counterparts, allowing for more compact designs. How that would work out in practice, who can say? One way or another, though, the curtain seems to have risen on nanotechnology's second act.
Curious about the world? To enjoy our mind-expanding science coverage, sign up to Simply Science, our weekly subscriber-only newsletter.