genetic mutations – NIH Director's Blog (original) (raw)

AI Model Takes New Approach to Performing Diagnostic Tasks in Multiple Cancer Types

Posted on October 3rd, 2024 by Dr. Monica M. Bertagnolli

Credit: Donny Bliss/NIH, Adobe Stock

In recent years, medical researchers have been looking for ways to use artificial intelligence (AI) technology for diagnosing cancer. So far, most AI models have been developed to perform specific tasks in cancer diagnosis, such as detecting cancer presence or predicting a tumor’s genetic profile in certain cancer types. But what if an AI system could be more flexible, like a large language model such as ChatGPT, performing a variety of diagnostic tasks across multiple cancer types?

As reported in the journal Nature, researchers have developed an AI system that can perform a wide range of cancer evaluation tasks and outperforms current AI methods in tasks like cancer cell detection and tumor origin identification. It was tested on 19 cancer types, leading the researchers to refer to it as “ChatGPT-like” in its flexibility. According to the research team, whose work is supported in part by NIH, this is also the first AI model based on analyzing slide images to not only accurately predict if a cancer is likely to respond to treatment, but also to validate these predictions across multiple patient groups around the world.

Today, when doctors order a biopsy to find out if cancer is present, those samples are sent to a pathologist, who examines the tissues or cells under a microscope to determine if they are cancerous. The team behind this AI model, led by Kun-Hsing Yu, Harvard Medical School, Boston, recognized that pathologists must analyze a wide variety of disease samples. To make accurate diagnoses in different cancer types, they must take many subtle factors into account.

Most earlier attempts to devise an AI model to analyze tissue samples have depended on training computers to recognize one cancer type at a time. In the new work, the researchers developed a more general-purpose pathology AI system that could analyze a broader range of tissues and sample types. To develop their Clinical Histopathology Imaging Evaluation Foundation (CHIEF) model, the researchers used an AI approach known as self-supervised learning. In this method, a computer is given large volumes of data, in this case 15 million pathology images, to allow it to identify intrinsic patterns and structures. This process allows a computer to “learn” from experience to identify informative features in a vast data set.

The tool was then trained further on more than 60,500 whole-slide images in tissues collected from 19 different parts of the body—such as the lungs, breast, prostate, kidney, brain, and bladder—to bolster the model’s ability to capture similarities and differences among cancer types. This training data was in part comprised of data from The Cancer Genome Atlas (TCGA) program and the Genotype-Tissue Expression (GTEx) Project, both NIH-supported resources. The researchers directed the model to consider both the image as a whole and its finer details, enabling it to interpret the image in a broader context than one region. They then put CHIEF to the test, using another 19,491 whole-slide images from 32 independent slide sets collected from 24 hospitals around the world.

They found that CHIEF worked equally well no matter how the samples were collected (biopsy or surgical excision) and in different clinical settings. In addition to detecting cancers and predicting a cancer’s tissue of origin, CHIEF also predicted with 70% accuracy whether a tissue carried one among dozens of genetic mutations that are commonly seen in cancers. CHIEF showed an ability to predict whether a sample contained mutated copies of 18 genes that oncologists use to make treatment decisions. CHIEF could predict better than earlier models how long a patient was likely to survive following a cancer diagnosis and how aggressively a particular cancer would grow.

This is all good news, but there’s much more work ahead before an AI model like this could be used in the clinic. Next steps for the researchers include training the model on images of tissues from rare cancers, as well as from pre-cancerous and non-cancerous conditions. With continued development and validation, the researchers aim to enable the system to identify cancers most likely to benefit from targeted or experimental therapies in hopes of improving outcomes for more people with cancer in diverse clinical settings around the world.

Reference:

Wang X, et al. A pathology foundation model for cancer diagnosis and prognosis prediction. Nature. DOI: 10.1038/s41586-024-07894-z (2024).

NIH Support: National Institute of General Medical Sciences, National Cancer Institute

Posted In: Health, Science, technology

Tags: AI, artificial intelligence, cancer, cancer diagnosis, computer learning, genetic mutations, genetics, imaging, pathology, technology

Finding New Genetic Mutations Amid Healthy Cells

Posted on October 3rd, 2019 by Dr. Francis Collins

Po-Ru Loh Courtesy of Loh Lab

You might recall learning in biology class that the cells constantly replicating and dividing in our bodies all carry the same DNA, inherited in equal parts from each parent. But it’s become increasingly clear in recent years that even seemingly healthy tissues contain neighborhoods of cells bearing their own acquired genetic mutations. The question is: What do all those altered cells mean for our health?

With support from a 2018 NIH Director’s New Innovator Award, Po-Ru Loh, Harvard Medical School, Boston, is on a quest to find out, though without the need for sequencing lots of DNA in his own lab. Loh will instead develop ultrasensitive computational tools to pick up on those often-subtle alterations within the vast troves of genomic data already stored in databases around the world.

How is that possible? The math behind it might be complex, but the underlying idea is surprisingly simple. His algorithms look for spots in the genome where a slight imbalance exists in the quantity of DNA inherited from mom versus dad.

Actually, Loh can’t tell from the data which parent provided any snippet of chromosomal DNA. But looking at DNA sequenced from a mixture of many cells, he can infer which stretches of DNA were most likely inherited together from a single parent.

Any slight skew in those quantities point the way to genomic territory where a tiny portion of chromosomal DNA either went missing or became duplicated in some cells. This common occurrence, especially in older adults, leads to a condition called genetic mosaicism, meaning that, contrary to most biology textbooks, all cells aren’t exactly the same.

By detecting those subtle imbalances in the data, Loh can pinpoint small DNA alterations, even when they occur in 1 in 1,000 cells collected from a person’s bloodstream, saliva, or tissues. That’s the kind of sensitivity that most scientists would not have thought possible.

Loh has already begun putting his new computational approach to work, as reported in Nature last year [1]. In DNA data from blood samples of more than 150,000 participants in the United Kingdom Biobank, his method uncovered well over 8,000 mosaic chromosomal alterations.

The study showed that some of those alterations were associated with an increased risk of developing blood cancers. However, it’s important to note that most people with evidence of mosaicism won’t go on to develop cancer. The researchers also made the unexpected discovery that some individuals carried genetic variants that made them more prone than others to pick up new mutations in their blood cells.

What’s especially exciting is Loh’s computational tools now make it possible to search for signs of mosaicism within all the genetic data that’s ever been generated. Even more importantly, these tools will allow Loh and other researchers to ask and answer important questions about the consequences of mosaicism for a wide range of diseases.

Reference:

[1] Insights into clonal haematopoiesis from 8,342 mosaic chromosomal alterations. Loh PR, Genovese G, Handsaker RE, Finucane HK, Reshef YA, Palamara PF, Birmann BM, Talkowski ME, Bakhoum SF, McCarroll SA, Price AL. Nature. 2018 Jul;559(7714):350-355.

Links:

Loh Lab (Harvard Medical School, Boston)

Loh Project Information (NIH RePORTER)

NIH Director’s New Innovator Award (Common Fund)

NIH Support: Common Fund; National Institute of Environmental Health Sciences

Posted In: Creative Minds

Tags: 2018 NIH Director’s New Innovator Award, blood cancer, cancer, chromosome, computational biology, DNA, genetic mosaicism, genetic mutations, mosaicism, mutations, United Kingdom Biobank

Study Finds Genetic Mutations in Healthy Human Tissues

Posted on June 18th, 2019 by Dr. Francis Collins

The standard view of biology is that every normal cell copies its DNA instruction book with complete accuracy every time it divides. And thus, with a few exceptions like the immune system, cells in normal, healthy tissue continue to contain exactly the same genome sequence as was present in the initial single-cell embryo that gave rise to that individual. But new evidence suggests it may be time to revise that view.

By analyzing genetic information collected throughout the bodies of nearly 500 different individuals, researchers discovered that almost all had some seemingly healthy tissue that contained pockets of cells bearing particular genetic mutations. Some even harbored mutations in genes linked to cancer. The findings suggest that nearly all of us are walking around with genetic mutations within various parts of our bodies that, under certain circumstances, may have the potential to give rise to cancer or other health conditions.

Efforts such as NIH’s The Cancer Genome Atlas (TCGA) have extensively characterized the many molecular and genomic alterations underlying various types of cancer. But it has remained difficult to pinpoint the precise sequence of events that lead to cancer, and there are hints that so-called normal tissues, including blood and skin, might contain a surprising number of mutations —perhaps starting down a path that would eventually lead to trouble.

In the study published in Science, a team from the Broad Institute at MIT and Harvard, led by Gad Getz and postdoctoral fellow Keren Yizhak, along with colleagues from Massachusetts General Hospital, decided to take a closer look. They turned their attention to the NIH’s Genotype-Tissue Expression (GTEx) project.

The GTEx is a comprehensive public resource that shows how genes are expressed and controlled differently in various tissues throughout the body. To capture those important differences, GTEx researchers analyzed messenger RNA sequences within thousands of healthy tissue samples collected from people who died of causes other than cancer.

Getz, Yizhak, and colleagues wanted to use that extensive RNA data in another way: to detect mutations that had arisen in the DNA genomes of cells within those tissues. To do it, they devised a method for comparing those tissue-derived RNA samples to the matched normal DNA. They call the new method RNA-MuTect.

All told, the researchers analyzed RNA sequences from 29 tissues, including heart, stomach, pancreas, and fat, and matched DNA from 488 individuals in the GTEx database. Those analyses showed that the vast majority of people—a whopping 95 percent—had one or more tissues with pockets of cells carrying new genetic mutations.

While many of those genetic mutations are most likely harmless, some have known links to cancer. The data show that genetic mutations arise most often in the skin, esophagus, and lung tissues. This suggests that exposure to environmental elements—such as air pollution in the lung, carcinogenic dietary substances in the esophagus, or the ultraviolet radiation in sunlight that hits the skin—may play important roles in causing genetic mutations in different parts of the body.

The findings clearly show that, even within normal tissues, the DNA in the cells of our bodies isn’t perfectly identical. Rather, mutations constantly arise, and that makes our cells more of a mosaic of different mutational events. Sometimes those altered cells may have a subtle growth advantage, and thus continue dividing to form larger groups of cells with slightly changed genomic profiles. In other cases, those altered cells may remain in small numbers or perhaps even disappear.

It’s not yet clear to what extent such pockets of altered cells may put people at greater risk for developing cancer down the road. But the presence of these genetic mutations does have potentially important implications for early cancer detection. For instance, it may be difficult to distinguish mutations that are truly red flags for cancer from those that are harmless and part of a new idea of what’s “normal.”

To further explore such questions, it will be useful to study the evolution of normal mutations in healthy human tissues over time. It’s worth noting that so far, the researchers have only detected these mutations in large populations of cells. As the technology advances, it will be interesting to explore such questions at the higher resolution of single cells.

Getz’s team will continue to pursue such questions, in part via participation in the recently launched NIH Pre-Cancer Atlas. It is designed to explore and characterize pre-malignant human tumors comprehensively. While considerable progress has been made in studying cancer and other chronic diseases, it’s clear we still have much to learn about the origins and development of illness to build better tools for early detection and control.

Reference:

[1] RNA sequence analysis reveals macroscopic somatic clonal expansion across normal tissues. Yizhak K, Aguet F, Kim J, Hess JM, Kübler K, Grimsby J, Frazer R, Zhang H, Haradhvala NJ, Rosebrock D, Livitz D, Li X, Arich-Landkof E, Shoresh N, Stewart C, Segrè AV, Branton PA, Polak P, Ardlie KG, Getz G. Science. 2019 Jun 7;364(6444).

Links:

Genotype-Tissue Expression Program

The Cancer Genome Atlas (National Cancer Institute/NIH)

Pre-Cancer Atlas (National Cancer Institute/NIH)

Getz Lab (Broad Institute, Cambridge, MA)

NIH Support: Common Fund; National Heart, Lung, and Blood Institute; National Human Genome Research Institute; National Institute of Mental Health; National Cancer Institute; National Library of Medicine; National Institute on Drug Abuse; National Institute of Neurological Diseases and Stroke

Posted In: News

Tags: cancer, DNA, esophagus, genetic mutations, genomics, GTEx, lungs, mosaic, mutations, normal tissues, Pre-Cancer Atlas, precancerous lesions, RNA, RNA-MuTect, skin, TCGA, The Cancer Genome Atlas, The Genotype-Tissue Expression Project

Yeast Reveals New Drug Target for Parkinson’s

Posted on November 5th, 2013 by Dr. Francis Collins

Caption: Left, yeast sick with too much α-synuclein, a protein that is implicated in Parkinson’s disease. Right, the same yeast cells after a dose of NAB, which seems to reverse the toxic effects of α-synuclein.
Credit: Daniel Tardiff, Whitehead Institute

Many progressive neurodegenerative disorders like Alzheimer’s, Huntington’s, and Parkinson’s disease, are characterized by abnormal clumps of proteins that clog up the cell and disrupt normal cellular functions. But it’s difficult to study these complex disease processes directly in the brain—so NIH-funded researchers, led by a team at the Whitehead Institute for Biomedical Research, Cambridge, MA, have turned to yeast for help.

Now, it may sound odd to study a brain disease in yeast, a microorganism long used in baking and brewing. After all, the brain is made up of billions of cells of many different types, while yeast grows as a single cell. But because the processes of protein production are generally conserved from yeast to humans, we can use this infinitely simpler organism to figure out what the proteins clumps are doing and test various drug candidates to halt the damage.

Posted In: Science

Tags: a-syn, alpha-synuclein, Alzheimer’s disease, brain, drug candidates, genetic mutations, Huntington's disease, induced Pluripotent Stem cells, N-aryl benzimidazole, NAB, National Institutes of Health, neurological diseases, neurons, NIH, Parkinson's disease, protein, protein clumping, Rsp5, ubiquitin tag, yeast