This is an analysis on an unprecedented scale. They studied over 140,000 versions of the Aβ42 peptide, which forms harmful plaques in the brain. It is the first map to reveal how mutations affect a protein in its transition state — a fleeting phase that is difficult to study. This finding opens up new avenues for preventing Alzheimer’s disease and suggests a method that can be applied to studying other proteins involved in different pathologies. The study, published in Science Advances, is a collaboration between the Wellcome Sanger Institute in the United Kingdom, the Institute for Bioengineering of Catalonia, and the Centre for Genomic Regulation in Barcelona.
A new large-scale study has mapped the first molecular events that drive the formation of harmful amyloid protein aggregates found in Alzheimer’s disease, pointing towards a new potential therapeutic target.
The scale at which we analysed the amyloid peptides was unprecedented – it’s something that hasn’t been done before and we have shown it’s a powerful new method to take forward.
Ben Lehner
Published today (11 June) in Science Advances, researchers from the Wellcome Sanger Institute, the Institute for Bioengineering of Catalonia (IBEC) and the Centre of Genomic Regulation (CRG) used large-scale genomics and machine learning to study over 140,000 versions of a peptide called Aβ42, which forms harmful plaques in the brain and is known to play a central role in Alzheimer’s disease.
“The approach we used in this study opens the door to revealing the structures of other protein transition states, including those implicated in other neurodegenerative diseases. The scale at which we analysed the amyloid peptides was unprecedented – it’s something that hasn’t been done before and we have shown it’s a powerful new method to take forward”, said Professor Ben Lehner, co-senior author, Head of Generative and Synthetic Genomics at the Wellcome Sanger Institute and ICREA Research Professor at the CRG. “We hope this takes us one step closer to developing treatments against Alzheimer’s disease and other neurodegenerative conditions.”
This research is a significant step towards helping scientists find new ways to prevent Alzheimer’s disease, and the methods used in the study could be applied widely to other protein reactions.
Understanding the origins of neurodegeneration
Over 55 million people are impacted by dementia globally and it is estimated that 60 to 70 per cent of these cases are Alzheimer’s disease.[1] Most current treatments for Alzheimer’s do not slow or stop the disease but help manage symptoms.
Amyloid beta (Aβ) is a peptide – a short chain of amino acids. Amyloid beta peptides have a tendency to clump and aggregate, forming elongated structures known as amyloid fibrils. Over time, these fibrils accumulate into plaques which are the pathological hallmarks of more than 50 neurodegenerative diseases, and most notably play a critical central role in Alzheimer’s disease.[2]
Our method of study is crucial to understand the first events in the process of protein aggregation that leads to dementia, but it also offers a powerful framework to dissect the key initiating steps of many biological reactions, not just those we’ve studied so far.
Benedetta Bolognesi
For free-flowing Aβ peptides to convert into stable, structured fibrils, they require a certain amount of energy. The intermediate, short-lived state right before the peptides begin to form a fibril is known as the ‘transition state’ – it is extremely unlikely to form, which is why fibrils never form in most people.
Understanding these structures and reactions is essential to developing therapies that could treat and prevent neurodegenerative diseases. However, it is very difficult to study short-lived high energy transition states using classical methods. As such, understanding how Aβ starts aggregating remains a major challenge in Alzheimer’s research.
Therefore, in this new study researchers from the Sanger Institute, Centre of Genomic Regulation and the Institute of Bioengineering of Catalonia sought to understand how changing the genetics of Aβ affects the rate of the aggregation reaction. Specifically, the researchers looked at Aβ42 – a type of Aβ peptide with 42 amino acids commonly found in those with Alzheimer’s.
“Our method of study is crucial to understand the first events in the process of protein aggregation that leads to dementia, but it also offers a powerful framework to dissect the key initiating steps of many biological reactions, not just those we’ve studied so far. I look forward to seeing all the ways in which this strategy will be employed in the future.”, stated Dr Benedetta Bolognesi, co-senior author and Group Leader of the Protein Phase Transitions in Health and Disease group at IBEC.
A large-scale analysis
The researchers used a combination of three techniques in order to handle large amounts of information about Aβ42 at the same time. The team used massively parallel DNA synthesis to study how changing amino acids in Aβ affects the amount of energy needed to form a fibril, and genetically engineered yeast cells to measure this rate of reaction.[3] They then used machine learning, a type of artificial intelligence,[4] to analyse the results and generate a complete energy landscape of amyloid beta aggregation reaction, showing the effect of all possible mutations in this protein on how fast fibrils are formed.
These techniques enabled the researchers to conduct the study at a large scale: “We measured the effect of more than 140,000 Aβ42 mutations and could apply neural networks, a type of machine learning, to extract, for each of them, the energy that drives the process of pathological aggregation.”, detailed Dr Mireia Seuma, co-first author formerly at IBEC and CRG, and now Senior Scientist at ALLOX. This scale has not been achieved before and helps improve the quality and accuracy of the models developed in the study. “Mutations, and the interactions between them, made it possible for us to “draw a portrait” of the transition state of the AB42 aggregation reaction. This is the key conformation driving the aggregation reaction, and it is extremely challenging (if not impossible) to study by classical biophysical methods.”, adds Seuma.
The researchers discovered that only a few key interactions between specific parts of the amyloid protein had a strong influence on the speed of fibril formation. They found that the Aβ42 aggregation reaction begins at the end of the protein, known as the C-terminal region, one of the hydrophobic cores of the protein – the tightly packed water-repellent region of the peptide. As it is here where the peptide starts aggregating into a fibril, the researchers suggest that it is the interactions in the C-terminal region that need to be prevented to protect against and treat Alzheimer’s disease.
This is the first large-scale map of how mutations influence a protein’s behaviour in the notoriously difficult to study transition state. By identifying the interactions that drive the formation of amyloid fibrils, the team believes that preventing the formation of this transition state could pave the way for new therapeutic strategies, offering hope for future Alzheimer’s treatments. Additionally, the researchers emphasise the wide usability of their method, noting it has potential to be used across a range of proteins and diseases in future studies.
Referenced paper:
Anna Arutyunyan, Mireia Seuma, Andre J. Faure, Benedetta Bolognesi, Ben Lehner. Massively parallel genetic perturbation suggests the energetic structure of an amyloid-β transition state. Science Advances (2025). DOI: 10.1126/sciadv.adv1422
[1] World Health Organization. Dementia. Available at: https://www.who.int/news-room/fact-sheets/detail/dementia [Last accessed: June 2025]
[2] C. M. Dobson (2017) The Amyloid Phenomenon and Its Links with Human Disease. Cold Spring Harb. Perspect. Biol. 9
[3] Massively parallel sequencing (MPS), also known as next-generation sequencing (NGS), is a high-throughput method for sequencing DNA or RNA fragments in parallel. It enables sequencing millions of fragments simultaneously.
[4] Machine learning involves using algorithms to analyse massive amounts of data from large datasets in order to identify patterns in how amino acid changes affect fibril formation and predict the energy landscape for amyloid beta aggregation.
