Biomedical research and AI: a synergy that can save lives

In recent years, Artificial Intelligence (AI) and its related technologies have revolutionised computational biology and medicine. Andrea Cavalli, Laboratory Director of the Institute for Research in Biomedicine (IRB), explained how these new tools are used in scientific research at the IRB in an article published by laRegione.

The first technology Andrea Cavalli refers to is AlphaFold, a system developed by DeepMind that can quickly and accurately predict the three-dimensional structure of proteins, which play a crucial role in our bodies. For this to happen, the sequence of amino acids that form the protein must be correct: "When even a single element is out of place, it can distort the entire structure, leading to potentially devastating effects, as seen in neurodegenerative genetic diseases like Parkinson's and Alzheimer's, as well as in tumours. For decades, understanding how proteins fold has been one of the most intriguing challenges in science," explained the IRB Laboratory Director.

Although existing methods could answer this question, AlphaFold has revolutionised the field by providing answers in a significantly shorter timeframe. This advancement opens up new areas of research, as noted by Andrea Cavalli: "In September 2023, the DeepMind team used AlphaFold to analyse the effect of 71 million harmful mutations in human proteins associated with rare genetic diseases and tumours. This work has mapped these mutations' structural and functional consequences on a large scale, creating a valuable resource for medical research and the development of new therapies."

AlphaFold has also been used within the IRB laboratories as part of a project in which Andrea Cavalli, Concetta Guerra and Jacopo Sgrignani took part: "We used AlphaFold to identify molecules capable of blocking the activity of the WWP2 protein with the aim of halting the uncontrolled growth of tumour cells in prostate cancer. Additionally, we are using this technology to design antibodies that target specific proteins known as 'modified' and hyperactive transcription factors, which are also involved in tumours and cannot be effectively studied using traditional experimental techniques."

The value of a technology such as AlphaFold was emphasised by the fact that in 2024, the Nobel Prize for Chemistry was awarded to Demis Hassabis, John Jumper and David Baker, founder of Rosetta, "an algorithm based on statistical and evolutionary models to predict the structure of proteins" which can be considered the ancestor of AlphaFold.

The second technology mentioned by Andrea Cavalli is Large Language Models (LLM), which allow for the rapid analysis of a large amount of information. "These advanced models have shown an impressive ability to learn complex relationships between objects and to "reason" about them in creative ways, similar to human thought. The impact of this technology extends beyond language, finding applications in various fields, including biology. In our laboratory, we are exploring how LLMs can interpret antibody and antigen sequences as if they were a language, decoding the intricate immunological alphabet. The ultimate goal of this project, called AntibodyGPT, is to develop an LLM capable of generating antibody sequences as 'sentences' in response to our 'prompts' to create neutralising antibodies against viral proteins or capable of blocking key proteins in autoimmune and oncological diseases," explained the IRB's Laboratory Director.

The next step in research involves the combined use of the two technologies: "The three-dimensional structure of proteins, revealed by AlphaFold, serves as a physical map of their functions and interactions. Meanwhile, the sequences of these proteins, analysed by large language models (LLMs), uncover the rules and meanings that guide complex molecular behaviour. This close interdependence allows for the integration of structural and sequential data, leading to a deeper and more detailed understanding of biomolecules and their interactions. As a result, this integration opens up new possibilities for biomedical innovation and enhances the effectiveness and creativity of personalised therapies," explained Andrea Cavalli. Furthermore, the two technologies open the way to the possibility of creating 'virtual twins' of patients, on which the effects of different therapies can be verified: "The fulfilment of this vision would transform personalised medicine, allowing advanced therapeutic simulations, optimisation treatments and acceleration of clinical research, reducing time and costs in the development of drugs" said Andrea Cavalli.

The IRB Laboratory Director concluded by recalling that there are numerous institutes of excellence in the Ticino area, such as the Swiss National Supercomputing Centre (CSCS), equipped with cutting-edge supercomputing infrastructure, Università della Svizzera italiana (USI), the IRB and the Institute of Oncology Research (IOR), who work together: "This synergy is driving science towards new frontiers, paving the way for predictive and personalised medicine. A collective commitment and a creative vision, between academia, industry and society, to realise the full potential of this revolution, transforming knowledge into concrete tools to improve human life."

The full article by Andrea Cavalli in laRegione is available at the following link. (Italian only)