Churchill Alumni Gabriel Ong recently had his first co-authored paper published and featured on the cover of ACS Synthetic Biology Journal. He tells us more about the paper, the process of getting it published, and why proteins fascinate him.
Give me a quick overview of how your paper came about?
“I was working as a Research Assistant in Dr Winston Koh’s lab at A*STAR in Singapore after graduating from Cambridge. My lab was focused on developing and deploying machine learning models for synthetic biology.
We were working on engineering a lot of proteins, and one of them was RNase A, which we were trying to use for medical diagnostics. As explained in the paper, you can edit your sample, clean it up, and then turn the protein off so it doesn’t go on to destroy what you’re actually trying to measure.”
What is it about this area that interests you?
“As early as high school, I was already quite fascinated by proteins. They seemed very interesting, these tiny machines in our body. There are famous ones are like ATP synthase, which is like a generator that spins. They literally keep our body working. I am very excited about harnessing these powerful machines to solve real-world problems, such as in sustainable chemical production or medical diagnostics. A lot of my past research experience has been related to proteins in some way. For instance, I did a science project in high school about incorporating aquaporins, which are the proteins responsible for filtering water in our kidneys, into water filtration membranes. Proteins are complicated and, to be honest, we still don’t understand a lot about them. Machine learning can provide us with a shortcut there, which was clearly demonstrated by the likes of AlphaFold and David Baker’s protein design group. These seminal works opened my eyes to what was possible at the intersection of protein science and computing, and drove me to pursue research in the area of machine learning for protein engineering.”
What was it like getting your first paper published?
“The experience of publishing was quite interesting, I learned a lot. I think when you read papers, in general, you don’t really appreciate the effort that went into writing it. And my supervisor (Dr Winston Koh) always talks about the “curse of knowledge”. It’s clear in your own mind, but writing it down for someone who has never seen it before and getting them to the same page as you is not trivial.
I had to kind of wipe my memory and then try to build it back up as I’m writing, so people can follow. In NatSci in Cambridge, they encourage us to draw diagrams. At the time, I didn’t fully appreciate why. Obviously some things are easier to explain in diagrams, but, once I was writing papers I realized being able to draw things can help you think about stuff, and is a useful tool for visual storytelling to help other people get on the same page as well. That’s when it clicked for me, why they were asking us to do that.
It was my first experience with peer review. The review process did take quite a while. It started in March or April 2025, and the paper only got published in December. We went through two rounds of peer review. I feel maybe we got lucky, but we didn’t really have any nasty reviewers. They gave constructive suggestions, suggested some experiments, or changes to the phrasing. It was nice to have them there as a counterbalance when we needed it.”
What are you doing now?
“I’m doing my PhD in San Diego, California, at the Scripps Research Institute, in computational biology.
I am currently rotating in Dr Stefano Forli’s lab, which made a computer program to find chemical compounds that can stick to proteins in the body – that’s how a lot of drugs work. It’s a challenge – you have to find the right shape, the right block, for the right hole. Except the shapes are even more complicated and maybe some sites don’t like each other.
I’m still working on something related to proteins, so that hasn’t changed. Currently I’m in a drug discovery lab, and we’re trying to bind chemicals to proteins, so now I’m learning more on the chemistry side. In particular, I want to learn more about the physics of the interactions between the proteins and chemicals, which is important because proteins do not exist in a vacuum, they are surrounded by an environment with other molecules.
Machine learning is great, but it’s very limited by the amount of data you have, and in science, there are a lot of things that we just don’t have a lot of data about, or it’s just not very good data, and I think that’s where physics comes in. Because physical laws are derived from first principles, they can constrain and guide model behavior even when empirical data is sparse or noisy. In a way it’s more efficient to come up with the underlying theories, but it’s often slower to calculate, so that’s the trade-off.
Even in the AI field, there’s a lot of movement towards models that actually have a physical understanding of the world. For example, in AI generated videos, there were a lot of problems initially, like too many fingers, or things floating away, because they don’t have an understanding of what a thing in this world is. A protein is also a physical thing, so it does make sense to better model how it how it behaves in the real physical world.
Furthermore, proteins are not rocks. Sometimes they’re moving, opening and closing. This has real implications for designing drugs based on protein structure. Say you need to get a drug to bind to a protein’s active site, you need to get it in when the protein’s open. You might need to open it yourself with the drug, prying it open. That’s a limitation in a lot of the current AI models for protein structure prediction: they only produce static structures, which is like trying to study how a horse gallops from a picture instead of a video.”
What next for you?
“To be honest, I’m not sure yet whether it will be academia or industry or something else. For me, what I am quite sure I want is just to be able to build something that’s my own. What label that gets, academia, startup, industry? I’m not sure yet. We’ll have to see. But creative intellectual control of what I’m doing is quite important to me, and there are many ways to get that.
I think statistically a lot of people end up doing something different from their PhD. I’m just working on the skills, both the technical skills and the soft skills, like communication, and then seeing where it goes from there, rather than being locked into a topic itself.
I think for me my North Star is still something to do with enzymes, with applications like degrading plastics or synthesizing complex chemicals. As I alluded to earlier, these are fascinating machines with many exciting applications. Even though I’m currently doing drug discovery, there’s still that part of the chemical binding to the protein. It doesn’t have the reaction part, but I think learning how to better model the interaction first will be a necessary first step before we can even talk about reactions, so that’s another reason why I’m in this field now.”
How did being at Churchill College help you?
“Churchill did have a very good environment, and things would have been very different if I wasn’t there. I was fortunate to be able to access funding through Churchill to do research placements with Prof Dame Janet Thornton and Dr Evangelia Petsalaki at the European Bioinformatics Institute. The latter placement was my first proper contact with machine learning and I was using it to try to predict the effects of mutations on proteins. I think that did lay the groundwork for my current interest, and it was a very good opportunity.
I was also very fortunate to have good mentorship from fellows at Churchill. Sonja Dunbar was my biology supervisor in first year, Adrian Barbrook was my supervisor in second year. Sonja and Adrian gave me a strong foundation in biology and biochemistry. Rita Monson supervised me and also gave feedback for research proposals for part 2 and 3. I could then start looking at interdisciplinary connections and taking things further into the computational space, like what I’m doing now. Katherine Stott was my DoS in the fourth year, and both Rita and Katherine gave me a lot of insight into the research process itself, and what it means. All of this was instrumental in my development as a scientist.
Churchill has a very large student population, and they do all kinds of different subjects. It is good to not just be in your bubble of scientists and talk to other people. I think informal conversation is also a good way to practice telling people what it is you actually do, at a level that’s easy for anyone to understand. Overall, I was very happy with my time at Churchill College.”
- Read Gabriel’s paper Design of a Labile RNase A Using Protein Language Models
- View Gabriel’s LinkedIn profile