Cerebral Systems associated with Mayo Clinic to create an AI genomic foundation model that predicts the best medical treatment 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 Cerebral Systemsin an interview with GamesBeat.

The Mayo Clinic, in collaboration with Cerebras Systems, announced significant progress in the development of artificial intelligence tools to improve patient care, today at the JP Morgan Healthcare Conference in San Francisco.

As part of Mayo Clinic’s commitment to transforming health care, the institution is leading the development of a world-class genomic foundation model, designed to support physicians and patients.

Like Nvidia and other semiconductor companies, Cerebras is focused on AI supercomputing. But its approach is very different from Nvidia, which relies on individual AI processors. Cerebras Systems designs an entire wafer — with multiple chips on a single silicon wafer — that collectively solves big AI problems and other computing tasks with much lower power consumption. Feldman says that it takes ten such systems to calculate the genomic foundation model in months of time. However, that takes much less time, effort, power and cost than traditional computational solutions, he said. PitchBook recently predicted that Cerebras will have an IPO in 2025.

Building on Mayo Clinic’s leadership in precision medicine, the model is designed to improve diagnostics and personalize treatment choices, with a primary focus on Rheumatoid Arthritis (RA). The treatment of RA presents a significant clinical challenge, often requiring extensive trials to find effective drugs for individual patients.

Traditional methods that examine a genetic marker have shown limited success in predicting response to treatment.

The combined team’s genomic model was trained by combining publicly available human reference genome data with Mayo’s comprehensive patient exome data. The human reference genome is a digital DNA sequence that represents a composite, “idealized” version of the human genome. It serves as a standard framework against which individual human genomes can be compared, enabling researchers to identify genetic differences.

In contrast to models trained only on the human reference genome, Mayo’s genomic foundation model showed better results in genomic variant classification because it was trained on data from 500 Mayo Clinic patients. As more patient data is included in the training, the team expects continued improvements in the quality of the model.

The team designed new benchmarks to determine the model’s clinically relevant capabilities, such as identifying specific medical conditions from DNA data, addressing a gap in available which are public benchmarks, which focus mainly on the identification of structural elements such as regulations or functional regions.

The Mayo Clinic Genomic Foundation Model demonstrates state-of-the-art accuracy in several key areas: 68-100% accuracy in RA benchmarks, 96% accuracy in cancer predisposing prediction, and 83% accuracy in cardiovascular phenotype prediction. These capabilities align with Mayo Clinic’s vision of delivering world-leading healthcare through AI technology. More testing needs to be done to confirm the results, Feldman said.

“Mayo Clinic is committed to using the most advanced AI technology to train models that will fundamentally change healthcare,” said Matthew Callstrom, Mayo Clinic’s medical director for strategy and chair of radiology. , in a statement. “Our collaboration with Cerebras has enabled us to develop a state-of-the-art AI model for genomics. In less than a year, we have developed great AI tools that help our doctors make more informed decisions based on genomic data.

“Mayo’s genomic foundation model sets a new bar for genomic models, surpassing not only standard tasks such as predicting the functional and regulatory properties of DNA but also enabling discoveries in complex correlations between genetic variants and medical conditions,” said Natalia Vassilieva, field CTO of Cerebras Sistema, in a statement. “Unlike current methods that focus on single variant association, this model allows the discovery of connections where collections of variants contribute to a particular condition.”

The rapid development of these models – often a multi-year effort – is facilitated by training Mayo Clinic’s custom models on the Cerebras AI platform. The Mayo Genomic Foundation Model represents significant 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.

AI development for chest X-ray

Separately, the Mayo Clinic today opened separate groundbreaking collaborations with Microsoft Research and Cerebras Systems in the field of generative artificial intelligence (AI), designed to personalize patient care, significantly speeding up time to diagnose and improve accuracy.

Announced during the JP Morgan Healthcare Conference, the projects focus on the development and testing of foundational models adapted for different applications, using the power of multimodal radiology images and data (including CT scan and MRI) with Microsoft Research and Cerebras genomic sequencing data.

Innovations have the potential to change how clinicians approach diagnosis and treatment, ultimately leading to better patient outcomes.

Foundation AI models are large, pre-trained models that can adapt and perform many tasks with little additional training. They learn from multiple datasets, gaining general knowledge that can be used in a variety of applications. This adaptability makes them efficient and versatile building blocks for many AI systems.

The Mayo Clinic and Microsoft Research are working together to develop foundational models that combine text and images. For this use case, Mayo and Microsoft Research collaborated to explore the use of generative AI in radiology using Microsoft Research’s AI technology and Mayo Clinic’s X-ray data.

Empowering clinicians with quick access to the information they need is central to this research project. The Mayo Clinic aims to develop a model that can automatically generate reports, evaluate tube placement and line chest X-rays, and identify changes from previous images. This proof-of-concept model seeks to improve clinician workflow and patient care by providing a more efficient and comprehensive analysis of radiographic images.

Mayo Clinic has 76,000 people and they see a lot of patients every year.

“We started a partnership to bring AI technology to health care. This allows us to combine their unique domain expertise, their unique data, with our AI expertise and our computing,” Feldman said.

He said that large-scale language models predict words, but genomic models predict nucleotides. If a nucleotide is broken by a mutation or transcription error, it can cause a disease or predict the onset of a disease.

Current models can only ask if flipping a single nucleotide predicts a disease. But Cerebras looked at flipping more than one nucleotide and created a more accurate model.

“What we’re using it for, together with the Mayo Clinic, is to predict which drug will work for a specific patient,” Feldman said.

It’s a billion-parameter foundation model, or 10 times bigger than AlphaFold, and it’s trained on a trillion tokens. That makes it more accurate, Feldman said.

Often, patients have to go through a process of trial-and-error to find out which medication works. But with this model, Feldman believes it can predict which drug will work for a specific person. The first target is rheumatoid arthritis, which afflicts 1.3 million Americans.

“While it’s still early, what we’ve been able to show is that we’re able to predict with impressive accuracy which drug will work for a patient,” he said.

In arthritis, the prediction accuracy is 87%. The data still need to be published and peer-reviewed.