TL;DR
Turn quiet afternoons into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
A multi-institutional research project called NATIVE-ID will study how proteins begin to aggregate, with the aim of training AI to predict early dysfunction linked to neurodegenerative disease. ARPA-H is providing up to $28.6 million for the broader effort; Randal Halfmann’s lab will receive about $4.1 million over two years to measure 50,000 proteins in yeast cells. Whether the resulting predictions will work in human neurons or lead to treatments remains to be established.
A new multi-institutional project will collect large-scale data on how proteins begin to clump together, with the aim of training artificial intelligence to predict early protein dysfunction linked to neurodegenerative diseases. The effort, called NATIVE-ID, is part of the U.S. Advanced Research Projects Agency for Health’s BIOGAMI program and will initially focus on frontotemporal lobar degeneration, or FTLD.
The project’s overall funding is up to $28.6 million, according to the project announcement. Randal Halfmann’s laboratory at the Stowers Institute for Medical Research in Kansas City has been selected to generate experimental data and is set to receive about $4.1 million over two years. The broader collaboration is led by the Innovative Genomics Institute at the University of California, Berkeley.
Halfmann’s lab plans to test 50,000 proteins in yeast cells under varied conditions intended to mimic changes that occur in human cells as they age. Using a technique called Distributed Amphifluoric FRET, or DAmFRET, the team expects to gather data from more than one million samples and produce over 10 billion measurements of protein aggregation. Those are projected outputs, not results already produced.
The data are intended to help AI models study intrinsically disordered proteins, which lack one stable shape and can shift among many configurations. The collaboration also includes scientists from Brown University, Emory University, Johns Hopkins University, Parallel Squared Technology Institute and Texas A&M University. Researchers at the Berkeley institute will contribute human neurons in which model predictions can ultimately be tested, according to the project description.
Building Data for Early Prediction
Many existing AI advances in biology focus on predicting the structure of proteins with relatively stable shapes. The project addresses a different challenge: intrinsically disordered proteins can take multiple forms, and their interactions may contribute to harmful aggregation. The research team says there is not yet a comparable, large-scale dataset for training models to interpret those behaviors.
If the approach works, researchers could gain a way to identify protein changes earlier in disease processes and to prioritize possible targets for further study. Halfmann said better predictions of disease probability and onset age could help more people seek preventive or early-stage treatment or enroll in clinical trials. That is a proposed future benefit, not a demonstrated outcome of the project.
The work may matter beyond FTLD because the project aims to develop methods applicable to other illnesses involving protein misfolding. FTLD was chosen as the initial focus partly because it shares genetic and biological features with amyotrophic lateral sclerosis, or ALS. The project’s results will need validation before they can support predictions about individual patients or clinical decisions.
As an affiliate, we earn on qualifying purchases.
From Pilot Studies to 50,000 Proteins
Proteins are made from amino-acid sequences, and those sequences influence how proteins behave even when they do not settle into a single fixed structure. Some proteins can aggregate, or stick together. Such aggregation is associated with neurodegenerative diseases including Alzheimer’s, Parkinson’s, ALS and Huntington’s, but the project announcement does not establish that aggregation alone causes these diseases.
Halfmann’s team developed DAmFRET in 2018 to measure protein self-assembly inside individual living cells. The lab has previously studied disease-related proteins in yeast; the announcement says those experiments can inform how the proteins behave in human cells. In 2023, the lab reported experimentally determining the structure of an initiating step in amyloid formation associated with Huntington’s disease.
Halfmann described that earlier work and studies of TDP-43, a protein associated with ALS and FTLD, as pilot efforts for NATIVE-ID. The new project will expand from hundreds of protein sequences in those studies to 50,000 proteins. Its scale is intended to give AI models substantially more experimental data than has previously been available for this problem.
“If we can better predict the probabilities and onset ages of disease, it could allow many more people to seek preventive or early-stage treatments or enroll in clinical trials.”
— Randal Halfmann, Stowers Institute investigator
neurodegenerative disease research tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Predictions Still Need Human Testing
The announced measurements have not yet been generated, and the project has not reported an AI model or a validated prediction. It remains unclear how accurately a model trained on yeast-cell experiments will forecast protein behavior in human neurons, or whether it will predict when disease begins in people.
The announcement does not provide a detailed public timeline for the dataset, model development or validation milestones. Nor does it establish that the work will produce a treatment, prevent disease or improve clinical outcomes. Those possibilities depend on results from the research and subsequent testing.
As an affiliate, we earn on qualifying purchases.
Data Collection and Neuron Validation
Halfmann’s lab is expected to begin the large-scale experimental work under its two-year award, measuring aggregation across the planned protein set and conditions. The research team’s stated goal is to use those results to train AI models, then test whether their predictions hold in human neurons provided by collaborators at the Innovative Genomics Institute.
Further updates will be needed to show whether the dataset reaches its projected scale, how well the models perform in human-cell tests and whether findings from the initial FTLD focus extend to other neurodegenerative diseases. No clinical application or treatment milestone was announced.
cell culture for neurodegeneration studies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the NATIVE-ID project?
NATIVE-ID is a multi-institutional research effort within ARPA-H’s BIOGAMI program. It aims to collect data on protein aggregation and use those data to train AI models to predict protein dysfunction.
How much funding has been announced?
The broader project is described as receiving up to $28.6 million. Randal Halfmann’s Stowers Institute lab is set to receive about $4.1 million over two years for its experimental work.
Which disease will the project study first?
The group will initially focus on frontotemporal lobar degeneration (FTLD), which shares genetic and biological features with ALS. The longer-term aim is to develop approaches relevant to other diseases involving protein misfolding.
Will the project predict disease in individual people?
That has not been demonstrated. The project aims to develop and test AI predictions about protein behavior, including in human neurons. The announcement does not report a validated tool for predicting an individual’s disease risk or onset age.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
