Professor José Rafael Penadés works at the Department of Infectious Disease at Imperial College London and he is also a lecturer at CEU UCH. He leads a team which has tested Google’s innovative AI co-scientist system, finding that it can generate valuable new scientific hypotheses.
Professor José Rafael Penadés works at the Department of Infectious Disease at Imperial College London and he is also a lecturer at the CEU Cardenal Herrera University in Valencia. He leads a team which has tested the ‘AI co-scientist’ system which uses artificial intelligence and which Google has just launched worldwide. This innovative AI system has been designed to work as a researcher would, generating highly valuable scientific hypotheses and accelerating the scientific discovery process.
The results have been surprising – and very positive. “An AI-generated hypothesis can take years to validate, ” said Professor Penadés. “So, when Google asked us to test its AI platform, we decided to task it with a question that we already knew the answer to through extensive, but unpublished, experimental work. Our research is about to be published in one of the most high-impact journals in the world, but at the time it was unpublished and unknown to the research community. Google’s AI co-scientist system put forward five hypotheses in response to our question and the most plausible of them coincided exactly with our own hypothesis, which we have now confirmed experimentally.
Professor Penadés has seen how Google’s new AI co-scientist platform is able to correctly analyse all the published research, formulate hypotheses and suggest experiments to validate them, all in record time. “The AI system able to put forward a hypothesis which we now know is right, but also one that took us years to understand, but it has also put forward other hypotheses which we are now working on,” Professor Penadés said.
“In just two days, Google’s AI co-scientist platform has formulated the same hypothesis which took us seven years of hard work to validate. We are convinced that this platform will enormously accelerate advances in science, transforming the way we generate and test scientific hypotheses.”
Validated by unpublished research
The research used to test the potential of Google’s AI co-scientist platform has found a new mechanism for gene transfer between bacteria. Professor Penadés‘s research group had observed the existence of a family of satellite viruses in different bacterial species. After several years of research, the group worked out the mechanism behind this phenomenon: these satellites were able to form tails from different viruses, enabling them to infect different bacterial species. Depending on the tail acquired, the satellites can go to one species or another.
Having established the existence of this mechanism, Prof Penadés’s team asked AI co-scientist about possible mechanisms which could explain the presence of identical satellites in several species: its answer was that the ability of these satellites to interact with different tails with different bacterial tropisms should be considered. “When we got the results, I went into some kind of shock. The only thing I could think was that the system had somehow accessed my computer. Not only because the first hypothesis was correct, but also because it was suggesting other ideas that we’d never even thought about,” said Prof Penadés.
The first idea that Google’s AI platform suggested coincided completely with the hypothesis his group had been working on for years and which had recently been experimentally validated. This finding, which the group are are about to publish in one of the most prestigious journals in the world, could not have been known by the AI system beforehand.
Potential to drive science forward
Prof Penadés’s research team, with members from Imperial College and CEU UCH, has found that these gene transfer pathways between bacteria of different species are essential in order to understand the evolutionary capacity of bacteria and how the mechanisms involved can make them more virulent and resistant to antibiotics – one of healthcare’s greatest challenges across the world.
“It’s true that Google’s AI co-scientist platform is still at an early stage of development, but we can see that it already has the potential to drive science forward. Given that there are urgent healthcare challenges for the whole world on the horizon, such as antimicrobial resistance, it’s clear that we need to do more with less and accelerate our discoveries. AI can help us to get the most out of the limited resources of our laboratories. It can be more than just a tool for us to use: it can act as an engine of creativity, accelerating the discovery process and reshaping the way in which we generate and test scientific hypotheses”, said Prof Penadés.
Publication of the research
The pre-print version of the study (i.e., it is still to undergo final checks before definitive publication) with which Google’s AI co-scientist system was tested is entitled “Chimeric infective particles expand species boundaries in phage inducible chromosomal island mobilization”, and it has been co-authored by researchers from Imperial College London, the CEU Cardenal Herrera University (CEU UCH) and the Instituto de Biomedicina de Valencia (IBV). The pre-print version can be seen here: https://doi.org/10.1101/2025.02.11.637232
Google has also published the results of the testing of its AI co-scientist system in an article entitled “Towards an AI co-scientist”, with contributions from researchers from Imperial College, the Fleming Initiative and Stanford University, amongst others.
