Join our daily and weekly newsletters for the latest updates and exclusive content covering cutting-edge AI. Learn more
Cerebras Systems has partnered with Mayo Clinic to create a baseline genomic AI model that predicts the best medical treatments for people with rheumatoid arthritis.
It could also be useful in predicting the best treatment for people with cancer and cardiovascular disease, said Andrew Feldman, CEO of Brain systemsin an interview with GamesBeat.
Mayo Clinic, in collaboration with Cerebras Systems, today announced significant advances in the development of artificial intelligence tools to advance patient care, at the JP Morgan Healthcare Conference in San Francisco.
As part of Mayo Clinic’s commitment to transforming health care, the institution led the development of a world-class genomics foundation model, designed to support physicians and patients.
Like Nvidia and other semiconductor companies, Cerebras focuses on AI supercomputing. But its approach is very different from Nvidia’s, which relies on individual AI processors. Cerebras Systems designs an entire wafer – with many chips on a single silicon wafer – that collectively solves big AI problems and other computing tasks with much lower power consumption. Feldman said it took dozens of such systems to calculate the basic genomic model over several months. Still, it represented far less time, effort, power and cost than traditional IT solutions, he said. PitchBook recently predicted that Cerebras would have an IPO in 2025.
Building on Mayo Clinic’s leadership in precision medicine, the model is designed to improve diagnostics and personalize treatment selection, with an initial focus on rheumatoid arthritis (RA). Treatment of RA presents a significant clinical challenge, often requiring multiple attempts to find effective medications for each patient.
Traditional approaches examining single genetic markers have shown limited success in predicting treatment response.
The joint team’s genomic model was trained by blending publicly available reference human genome data with Mayo’s patient whole exome data. The human reference genome is a digital DNA sequence representing an “idealized” composite version of the human genome. It serves as a standard framework to which individual human genomes can be compared, allowing researchers to identify genetic variations.
Unlike models trained exclusively on the human reference genome, the Mayo Genomic Baseline model demonstrates significantly better results in classifying genomic variants because it was trained on data from 500 Mayo Clinic patients. As more patient data is incorporated into training, the team expects continued improvement in model quality.
The team designed new criteria to evaluate the model’s clinically relevant capabilities, such as detecting specific medical conditions from DNA data, filling a gap in publicly available criteria, which primarily focus on identification structural elements such as regulatory or functional regions.
The Mayo Clinic Genomic Core Model demonstrates industry-leading accuracy in several key areas: 68-100% accuracy in RA benchmarks, 96% accuracy in predicting cancer predisposition, and 83% accuracy in predicting cardiovascular phenotype. These capabilities align with Mayo Clinic’s vision to deliver world-leading healthcare through AI technology. Additional testing will need to be done to verify the results, Feldman said.
“Mayo Clinic is committed to using the most advanced AI technology to train models that will fundamentally transform health care,” said Matthew Callstrom, Mayo Clinic medical director for strategy and chair of radiology. “Our collaboration with Cerebras has allowed us to create a cutting-edge AI model for genomics. In less than a year, we have developed promising AI tools that will help our doctors make more informed decisions based on genomic data.
“Mayo’s genomic core model sets a new bar for genomic models, excelling not only at standard tasks such as predicting the functional and regulatory properties of DNA, but also enabling the discovery of complex correlations between genetic variants and medical conditions,” said Natalia Vassilieva, Field CTO at Cerebras. Systems, in a press release. “Unlike current approaches focused on single-variant associations, this model enables the discovery of connections where collections of variants contribute to a particular condition.”
The rapid development of these models – typically a multi-year effort – has been accelerated by the training of Mayo Clinic’s custom models on the Cerebras AI platform. The Mayo Genomic Foundation model represents important steps toward improving clinical decision support and advancing precision medicine.
Cerebras’ flagship product is the CS-3, a system powered by the Wafer-Scale Engine-3.
Advancing AI for Chest X-rays
Separately, Mayo Clinic today unveiled separate groundbreaking collaborations with Microsoft Research and Cerebras Systems in generative artificial intelligence (AI), designed to personalize patient care, dramatically accelerate time to diagnosis, and improve precision.
Announced at the JP Morgan Healthcare Conference, the projects focus on developing and testing custom core models for various applications, leveraging the power of multimodal radiology images and data (including CT and MRI scans). ) with Microsoft Research and genomic sequencing data with Cerebras.
These innovations have the potential to transform the way clinicians approach diagnosis and treatment, leading to better patient outcomes.
Foundation AI models are large, pre-trained models that can scale and perform many tasks with minimal additional training. They learn from massive datasets and gain general knowledge that can be used in various applications. This adaptability makes them effective and versatile building blocks for many AI systems.
Mayo Clinic and Microsoft Research are collaboratively developing basic models that integrate text and images. For this use case, Mayo and Microsoft Research are working together to explore the use of generative AI in radiology using AI technology from Microsoft Research and radiology data from Mayo Clinic.
Giving clinicians instant access to the information they need is at the heart of this research project. Mayo Clinic aims to develop a model that can automatically generate reports, evaluate the placement of tubes and lines in chest X-rays, and detect changes from previous images. This proof-of-concept model aims to improve clinician workflow and patient care by providing more efficient and comprehensive analysis of radiographic images.
The Mayo Clinic has a staff of 76,000 and sees a large number of patients each year.
“We have begun a partnership to bring AI technology to healthcare. It allowed us to kind of combine their domain expertise, their remarkable data, with our AI expertise and our computation,” Feldman said.
He said large linguistic models predict words, but genomic models predict nucleotides. When a nucleotide is flipped during a mutation or transcription error, it could be the cause of disease or predict the onset of disease.
Existing models can only ask whether single-nucleotide flipping predicts disease. But Cerebras looks at flipping more than one nucleotide and comes up with a more precise model.
“We use it, in collaboration with the Mayo Clinic, to predict which drug will work for a specific patient,” Feldman said.
It is a base model with a billion parameters, 10 times larger than AlphaFold, and it was trained on a trillion tokens. That makes it more accurate, Feldman said.
Too often, patients must go through a process of trial and error to determine which medication will work. But with this model, Feldman believes he can predict which drug will work on a specific person. The first target is rheumatoid arthritis, which affects 1.3 million Americans.
“Although it is still early, what we were able to show was that we were able to predict with impressive accuracy which drug would work for a given patient,” he said.
For arthritis, the prediction accuracy was 87%. The data still needs to be published and peer-reviewed.
