bioinnovationbeacon.com

Author: admin

  • Researchers harness AI to repurpose existing drugs for treatment of rare diseases

    Researchers harness AI to repurpose existing drugs for treatment of rare diseases

    There are more than 7,000 rare and undiagnosed diseases globally.

    Although each condition occurs in a small number of individuals, collectively these diseases exert a staggering human and economic toll because they affect some 300 million people worldwide.

    Yet, with a mere 5 to 7 percent of these conditions having an FDA-approved drug, they remain largely untreated or undertreated.

    Developing new medicines represents a daunting challenge, but a new artificial intelligence tool can propel the discovery of new therapies from existing medicines, offering hope for patients with rare and neglected conditions and for the clinicians who treat them.

    The AI model, called TxGNN, is the first one developed specifically to identify drug candidates for rare diseases and conditions with no treatments.

    It identified drug candidates from existing medicines for more than 17,000 diseases, many of them without any existing treatments. This represents the largest number of diseases that any single AI model can handle to date. The researchers note that the model could be applied to even more diseases beyond the 17,000 it worked on in the initial experiments.

    The work, described Sept. 25 in Nature Medicine, was led by scientists at Harvard Medical School. The researchers have made the tool available for free and want to encourage clinician-scientists to use it in their search for new therapies, especially for conditions with no or with limited treatment options.

    “With this tool we aim to identify new therapies across the disease spectrum but when it comes to rare, ultrarare, and neglected conditions, we foresee this model could help close, or at least narrow, a gap that creates serious health disparities,” said lead researcher Marinka Zitnik, assistant professor of biomedical informatics in the Blavatnik Institute at HMS.

    “This is precisely where we see the promise of AI in reducing the global disease burden, in finding new uses for existing drugs, which is also a faster and more cost-effective way to develop therapies than designing new drugs from scratch,” added Zitnik, who is an associate faculty member at the Kempner Institute for the Study of Natural and Artificial Intelligence at Harvard University.

    The new tool has two central features — one that identifies treatment candidates along with possible side effects and another one that explains the rationale for the decision.

    In total, the tool identified drug candidates from nearly 8,000 medicines (both FDA-approved medicines and experimental ones now in clinical trials) for 17,080 diseases, including conditions with no available treatments. It also predicted which drugs would have side effects and contraindications for specific conditions — something that the current drug discovery approach identifies mostly by trial and error during early clinical trials focused on safety.

    Compared against the leading AI models for drug repurposing, the new tool was nearly 50 percent better, on average, at identifying drug candidates. It was also 35 percent more accurate in predicting what drugs would have contraindications.

    Advantages of using already approved drugs

    Repurposing existing drugs is an alluring way to develop new treatments because it relies on medicines that have been studied, have well-understood safety profiles, and have gone through the regulatory approval process.

    Most medicines have multiple effects beyond the specific targets they were originally developed and approved for. But many of these effects remain undiscovered and understudied during initial testing, clinical trials, and review, only emerging after years of use by millions of people. Indeed, nearly 30 percent of FDA- approved drugs have acquired at least one additional indication for treatment following initial approval, and many have acquired tens of additional treatment indications over the years.

    This approach to drug repurposing is haphazard at best. It relies on patient reports of unexpected beneficial side effects or on physicians’ intuition about whether to use a drug for a condition that it was not intended for, a practice known as off-label use.

    “We’ve tended to rely on luck and serendipity rather than on strategy, which limits drug discovery to diseases for which drugs already exist,” Zitnik said.

    The benefits of drug repurposing extend beyond diseases without treatments, Zitnik noted.

    “Even for more common diseases with approved treatments, new drugs could offer alternatives with fewer side effects or replace drugs that are ineffective for certain patients,” she said.

    What makes the new AI tool better than existing models

    Most current AI models used for drug discovery are trained on a single disease or a handful of conditions. Rather than focusing on specific diseases, the new tool was trained in a manner that enables it to use existing data to make new predictions. It does so by identifying shared features across multiple diseases, such as shared genomic aberrations.

    For example, the AI model pinpoints shared disease mechanisms based on common genomic underpinnings, which allows it to extrapolate from a well-understood disease with known treatments to a poorly understood one with no treatments.

    This capacity, the research team said, brings the AI tool closer to the type of a reasoning a human clinician might use to generate novel ideas if they had access to all the preexisting knowledge and raw data that the AI model does but that the human brain cannot possibly access or store.

    The tool was trained on vast amounts of data, including DNA information, cell signaling, levels of gene activity, clinical notes, and more. The researchers tested and refined the model by asking it to perform various tasks. Finally, the tool’s performance was validated on 1.2 million patient records and asked to identify drug candidates for various diseases.

    The researchers also asked the tool to predict what patient characteristics would render the identified drug candidates contraindicated for certain patient populations.

    Another task involved asking the tool to identify existing small molecules that might effectively block the activity of certain proteins implicated in disease-causing pathways and processes.

    In a test designed to gauge the model’s ability to reason as a human clinician might, the researchers prompted the model to find drugs for three rare conditions it had not seen as part of its training — a neurodevelopmental disorder, a connective-tissue disease, and a rare genetic condition that causes water imbalance.

    The researchers then compared the model’s recommendations for drug therapy against current medical knowledge about how the suggested drugs work. In every example, the tool’s recommendations aligned with current medical knowledge.

    Moreover, the model not only identified medicines for all three diseases but also provided the rationale behind its decision. This explainer feature allows for transparency and can increase physician confidence.

    The researchers caution that any therapies identified by the model would require additional evaluation for dosing and timing of delivery. But, they add, with this unprecedented capacity, the new AI model would expedite drug repurposing in a manner not possible until now. The team is already collaborating with several rare disease foundations to help identify possible treatments.

  • How do rare genetic variants affect health? AI provides more accurate predictions

    How do rare genetic variants affect health? AI provides more accurate predictions

    Whether we are predisposed to particular diseases depends to a large extent on the countless variants in our genome. However, particularly in the case of genetic variants that only rarely occur in the population, the influence on the presentation of certain pathological traits has so far been difficult to determine. Researchers from the German Cancer Research Center (DKFZ), the European Molecular Biology Laboratory (EMBL) and the Technical University of Munich have introduced an algorithm based on deep learning that can predict the effects of rare genetic variants. The method allows persons with high risk of disease to be distinguished more precisely and facilitates the identification of genes that are involved in the development of diseases.

    Every person’s genome differs from that of their fellow human beings in millions of individual building blocks. These differences in the genome are known as variants. Many of these variants are associated with particular biological traits and diseases. Such correlations are usually determined using so-called genome-wide association studies.

    But the influence of rare variants, which occur with a frequency of only 0.1% or less in the population, is often statistically overlooked in association studies. “Rare variants in particular often have a significantly greater influence on the presentation of a biological trait or a disease,” says Brian Clarke, one of the first authors of the present study. “They can therefore help to identify those genes that play a role in the development of a disease and that can then point us in the direction of new therapeutic approaches,” adds co-first author Eva Holtkamp.

    In order to better predict the effects of rare variants, teams led by Oliver Stegle and Brian Clarke at the DKFZ and EMBL and Julien Gagneur at the Technical University of Munich have now developed a risk assessment tool based on machine learning. “DeepRVAT” (rare variant association testing), as the researchers named the method, is the first to use artificial intelligence (AI) in genomic association studies to decipher rare genetic variants.

    The model was initially trained on the sequence data (exome sequences) of 161,000 individuals from the UK Biobank. In addition, the researchers fed in information on genetically influenced biological traits of the individual persons as well as on the genes involved in the traits. The sequences used for training comprised around 13 million variants. For each of these, detailed “annotations” are available, providing quantitative information on the possible effects that the respective variant can have on cellular processes or on the protein structure. These annotations were also a central component of the training.

    After training, DeepRVAT is able to predict for each individual which genes are impaired in their function by rare variants. To do this, the algorithm uses individual variants and their annotations to calculate a numerical value that describes the extent to which a gene is impaired and its potential impact on health.

    The researchers validated DeepRVAT on genome data from the UK Biobank. For 34 tested traits, i.e., disease-relevant blood test results, the testing method found 352 associations with genes involved, far outperforming all previously existing models. The results obtained with DeepRVAT proved to be very robust and better replicable in independent data than the results of alternative approaches.

    Another important application of DeepRVAT is the evaluation of genetic predisposition to certain diseases. The researchers combined DeepRVAT with polygenic risk scoring based on more common genetic variants. This significantly improved the accuracy of the predictions, especially for high-risk variants. In addition, it turned out that DeepRVAT recognized genetic correlations for numerous diseases — including various cardiovascular diseases, types of cancer, metabolic and neurological diseases — that had not been found with existing tests.

    “DeepRVAT has the potential to significantly advance personalized medicine. Our method functions regardless of the type of trait and can be flexibly combined with other testing methods,” says physicist and data scientist Oliver Stegle. His team now wants to further test the risk assessment tool in large-scale trials as quickly as possible and bring it into application. The scientists are already in contact with the organizers of INFORM, for example. The aim of this study is to use genomic data to identify individually tailored treatments for children with cancer who suffer a relapse. DeepRVAT could help to uncover the genetic basis of certain childhood cancers.

    “I find the potential impact of DeepRVAT on rare disease applications exciting. One of the major challenges in rare disease research is the lack of large-scale, systematic data. Leveraging the power of AI and the half a million exomes in the UK Biobank, we have objectively identified which genetic variants most significantly impair gene function,” says Julien Gagneur from the Technical University of Munich.

    The next step is to integrate DeepRVAT into the infrastructure of the German Human Genome Phenome Archive (GHGA) in order to facilitate applications in diagnostics and basic research. Another advantage of DeepRVAT is that the method requires significantly less computing power than comparable models. DeepRVAT is available as a user-friendly software package that can either be used with the pre-trained risk assessment models or trained with researchers’ own data sets for specialized purposes.

  • Researchers test ChatGPT, other AI models against real-world students

    Researchers test ChatGPT, other AI models against real-world students

    William Hersh, M.D., who has taught generations of medical and clinical informatics students at Oregon Health & Science University, found himself curious about the growing influence of artificial intelligence. He wondered how AI would perform in his own class.

    So, he decided to try an experiment.

    He tested six forms of generative, large-language AI models — for example ChatGPT — in an online version of his popular introductory course in biomedical and health informatics to see how they performed compared with living, thinking students. A study published in the journal npj Digital Medicine, revealed the answer: Better than as many as three-quarters of his human students.

    “This does raise concern about cheating, but there is a larger issue here,” Hersh said. “How do we know that our students are actually learning and mastering the knowledge and skills they need for their future professional work?”

    As a professor of medical informatics and clinical epidemiology in the OHSU School of Medicine, Hersh is especially attuned to new technologies. The role of technology in education is nothing new, Hersh said, recalling his own experience as a high school student in the 1970s during the transition from slide rules to calculators.

    Yet, the shift to generative AI represents an exponential leap forward.

    “Clearly, everyone should have some kind of foundation of knowledge in their field,” Hersh said. “What is the foundation of knowledge you expect people to have to be able to think critically?”

    Large-language models

    Hersh and co-author Kate Fultz Hollis, an OHSU informatician, pulled the knowledge assessment scores of 139 students who took the introductory course in biomedical and health informatics in 2023. They prompted six generative AI large language models with student assessment materials from the course. Depending on the model, AI scored in the top 50th to 75th percentile on multiple-choice questions that were used in quizzes and a final exam that required short written responses to questions.

    “The results of this study raise significant questions for the future of student assessment in most, if not all, academic disciplines,” the authors write.

    The study is the first to compare large-language models to students for a full academic course in the biomedical field. Hersh and Fultz Hollis noted that a knowledge-based course such as this one may be especially ripe for generative, large-language models, in contrast to more participatory academic courses that help students develop more complex skills and abilities.

    Hersh remembers his experience in medical school.

    “When I was a medical student, one of my attending physicians told me I needed to have all the knowledge in my head,” he said. “Even in the 1980s, that was a stretch. The knowledge base of medicine has long surpassed the capacity of the human brain to memorize it all.”

    Maintaining the human touch

    Yet, he believes there’s a fine line between making sensible use of technical resources to advance learning and over-reliance to the point that it inhibits learning. Ultimately, the goal of an academic health center like OHSU is to educate health care professionals capable of caring for patients and optimizing the use of data and information about them in the real world.

    In that sense, he said, medicine will always require the human touch.

    “There are a lot of things that health care professionals do that are pretty straightforward, but there are those instances where it gets more complicated and you have to make judgment calls,” he said. “That’s when it helps to have that broader perspective, without necessarily needing to have every last fact in your brain.”

    With fall classes starting soon, Hersh said he’s not worried about cheating.

    “I update the course each year,” he said. “In any scientific field, there are new advancements all the time and large-language models aren’t necessarily up to date on all of it. This just means we’ll have to look at newer or more nuanced tests where you won’t get the answer out of ChatGPT.”

  • Can chatbots help with genetic testing for cancer risk?

    Can chatbots help with genetic testing for cancer risk?

    In a study from Huntsman Cancer Institute at the University of Utah (the U) and NYU Langone Perlmutter Cancer Center, researchers found that a specialized chatbot can effectively assist patients in deciding whether to pursue genetic testing, offering an alternative to traditional genetic counseling.

    Kimberly Kaphingst, ScD, research director of the Genetic Counseling Shared Resource and co-leader of the Cancer Control and Population Sciences Program at Huntsman Cancer Institute, and professor of communication at the U, says the results of the BRIDGE (Broadening the Reach, Impact, and Delivery of Genetic Services) trial could help expand patient access to genetic care.

    According to the American Cancer Society, up to 10% of all cancers may be caused by inherited genetic changes.

    “There’s a substantial number of people who have an inherited cancer syndrome. The vast majority don’t know it,” says Kaphingst. “As we are getting better at recognizing people who are in need of genetic testing, we were very interested as a team in coming up with sustainable and scalable ways to actually provide those types of genetic services.”

    The current standard of care model for genetic testing involves a two-appointment process. Patients first meet with a genetic counselor for a pre-test appointment, during which they discuss their family history, as well as the risks, benefits, and limitations of testing. If patients choose to proceed with testing, they schedule a second appointment to analyze the results with the counselor.

    The BRIDGE trial used an algorithm to find patients at higher risk for inherited cancer syndromes based on their self-reported family health histories.

    Researchers then divided more than 3,000 Utah and New York participants into two groups — one pursuing the standard two-appointment model, and the other engaging with a chatbot designed and scripted to provide genetics education instead of having a pre-test appointment with a genetic counselor.

    Participants in the chatbot group were sent a message through MyChart, an online patient health portal, recommending genetic services and providing a link to launch a chatbot. They then received information about genetic testing and were able to ask questions to help them decide if they should proceed with testing.

    Researchers found outcomes between the two groups were equally likely to complete genetic testing, demonstrating that a chatbot is a viable alternative to the traditional model.

    “Our goal wasn’t to see if the chatbot was better or for worse. The question was, is this another model that we can use to offer genetic services to patients and have similar outcomes. For a lot of people, the chatbot model provided enough information,” says Kaphingst. “The chatbot can take some of the burden off genetic counselors and help provide genetic testing to more patients who are eligible.”

    Researcher and genetic counselor Rachelle Chambers, MS, CGC, manager of the High-Risk Cancer Genetics Program at NYU Langone Perlmutter Cancer Center, says the promising results of this equivalency trial are crucial as knowledge of and demand for genetic testing is on the rise.

    “Twenty years ago, we were testing for a handful of cancer-related genes, like BRCA1 and BRCA2, which can lead to an increased risk for breast and ovarian cancer. Now, we might analyze 100 different genes linked to cancers,” says Chambers. “From my perspective as a genetic counselor, there are not enough genetic specialists to meet the increased demand of patients that could benefit from this type of testing.”

    Patients who know they are genetically more likely to develop cancer can be proactive in prevention, through increased screenings, taking medications, pursuing surgery to reduce risk, and changing personal behaviors.

    The results of the BRIDGE trial have been published in JAMA Network Open.

    The trial was co-led by Saundra Buys, MD, Huntsman Cancer Institute investigator and professor in the department of medicine at the U, and Meenakshi Sigireddi, MD, assistant professor of medicine at NYU Grossman School of Medicine. Additional Huntsman Cancer Institute collaborators include Wendy Kohlmann, MS, now at the U.S. Department of Veterans Affairs, Sarah Colonna, MD, Whitney F. Espinel, CGC, MS, Amanda Gammon, MS, CGC, Josh Schiffman, MD, professor of pediatrics at the U, Kensaku Kawamoto, MD, PhD, MHS, FACMI, FAMIA, professor of clinical informatics at the U, Guilherme Del Fiol, MD, PhD, FACMI, professor of biomedical informatics at the U, and David Wetter, PhD, MS, professor of population health sciences at the U. Michael Flynn, MD, adjunct assistant professor at the U, and Rachel Hess, MD, MS, professor of population health sciences at the U, also contributed to the study.

    This study was supported by the National Institutes of Health/National Cancer Institute through U01CA232826, which was part of the Inherited Cancer Syndrome Collaborative, P30 CA02014; and Huntsman Cancer Foundation.

  • A new artificial intelligence tool for cancer

    A new artificial intelligence tool for cancer

    Scientists at Harvard Medical School have designed a versatile, ChatGPT-like AI model capable of performing an array of diagnostic tasks across multiple forms of cancers.

    The new AI system, described Sept. 4 in Nature, goes a step beyond many current AI approaches to cancer diagnosis, the researchers said. (DOI 10.1038/s41586-024-07894-z)

    Current AI systems are typically trained to perform specific tasks — such as detecting cancer presence or predicting a tumor’s genetic profile — and they tend to work only in a handful of cancer types. By contrast, the new model can perform a wide array of tasks and was tested on 19 cancer types, giving it a flexibility like that of large language models such as ChatGPT.

    While other foundation AI models for medical diagnosis based on pathology images have emerged recently, this is believed to be the first to predict patient outcomes and validate them across several international patient groups.

    “Our ambition was to create a nimble, versatile ChatGPT-like AI platform that can perform a broad range of cancer evaluation tasks,” said study senior author Kun-Hsing Yu, assistant professor of biomedical informatics in the Blavatnik Institute at Harvard Medical School. “Our model turned out to be very useful across multiple tasks related to cancer detection, prognosis, and treatment response across multiple cancers.”

    The AI model, which works by reading digital slides of tumor tissues, detects cancer cells and predicts a tumor’s molecular profile based on cellular features seen on the image with superior accuracy to most current AI systems. It can forecast patient survival across multiple cancer types and accurately pinpoint features in the tissue that surrounds a tumor — also known as the tumor microenvironment — that are related to a patient’s response to standard treatments, including surgery, chemotherapy, radiation, and immunotherapy. Finally, the team said, the tool appears capable of generating novel insights — it identified specific tumor characteristics previously not known to be linked to patient survival.

    The findings, the research team said, add to growing evidence that AI-powered approaches can enhance clinicians’ ability to evaluate cancers efficiently and accurately, including the identification of patients who might not respond well to standard cancer therapies.

    “If validated further and deployed widely, our approach, and approaches similar to ours, could identify early on cancer patients who may benefit from experimental treatments targeting certain molecular variations, a capability that is not uniformly available across the world,” Yu said.

    Training and performance

    The team’s latest work builds on Yu’s previous research in AI systems for the evaluation of colon cancer and brain tumors. These earlier studies demonstrated the feasibility of the approach within specific cancer types and specific tasks.

    The new model, called CHIEF (Clinical Histopathology Imaging Evaluation Foundation), was trained on 15 million unlabeled images chunked into sections of interest. The tool was then trained further on 60,000 whole-slide images of tissues including lung, breast, prostate, colorectal, stomach, esophageal, kidney, brain, liver, thyroid, pancreatic, cervical, uterine, ovarian, testicular, skin, soft tissue, adrenal gland, and bladder. Training the model to look both at specific sections of an image and the whole image allowed it to relate specific changes in one region to the overall context. This approach, the researchers said, enabled CHIEF to interpret an image more holistically by considering a broader context, instead of just focusing on a particular region.

    Following training, the team tested CHIEF’s performance on more than 19,400 whole-slide images from 32 independent datasets collected from 24 hospitals and patient cohorts across the globe.

    Overall, CHIEF outperformed other state-of-the-art AI methods by up to 36 percent on the following tasks: cancer cell detection, tumor origin identification, predicting patient outcomes, and identifying the presence of genes and DNA patterns related to treatment response.Because of its versatile training, CHIEF performed equally well no matter how the tumor cells were obtained — whether via biopsy or through surgical excision. And it was just as accurate, regardless of the technique used to digitize the cancer cell samples. This adaptability, the researchers said, renders CHIEF usable across different clinical settings and represents an important step beyond current models that tend to perform well only when reading tissues obtained through specific techniques.

    Cancer detection

    CHIEF achieved nearly 94 percent accuracy in cancer detection and significantly outperformed current AI approaches across 15 datasets containing 11 cancer types. In five biopsy datasets collected from independent cohorts, CHIEF achieved 96 percent accuracy across multiple cancer types including esophagus, stomach, colon, and prostate. When the researchers tested CHIEF on previously unseen slides from surgically removed tumors of the colon, lung, breast, endometrium, and cervix, the model performed with more than 90 percent accuracy.

    Predicting tumors’ molecular profiles

    A tumor’s genetic makeup holds critical clues to determine its future behavior and optimal treatments. To get this information, oncologists order DNA sequencing of tumor samples, but such detailed genomic profiling of cancer tissues is not done routinely nor uniformly across the world due to the cost and time involved in sending samples to specialized DNA sequencing labs. Even in well-resourced regions, the process could take several weeks. It’s a gap that AI could fill, Yu said.

    Quickly identifying cellular patterns on an image suggestive of specific genomic aberrations could offer a quick and cost-effective alternative to genomic sequencing, the researchers said.

    CHIEF outperformed current AI methods for predicting genomic variations in a tumor by looking at the microscopic slides. This new AI approach successfully identified features associated with several important genes related to cancer growth and suppression, and it predicted key genetic mutations related to how well a tumor might respond to various standard therapies. CHIEF also detected specific DNA patterns related to how well a colon tumor might respond to a form of immunotherapy called immune checkpoint blockade. When looking at whole-tissue images, CHIEF identified mutations in 54 commonly mutated cancer genes with an overall accuracy of more than 70 percent, outperforming the current state-of-the-art AI method for genomic cancer prediction. Its accuracy was greater for specific genes in specific cancer types.

    The team also tested CHIEF on its ability to predict mutations linked with response to FDA-approved targeted therapies across 18 genes spanning 15 anatomic sites. CHIEF attained high accuracy in multiple cancer types, including 96 percent in detecting a mutation in a gene called EZH2 common in a blood cancer called diffuse large B-cell lymphoma. It achieved 89 percent for BRAF gene mutation in thyroid cancer, and 91 percent for NTRK1 gene mutation in head and neck cancers.

    Predicting patient survival

    CHIEF successfully predicted patient survival based on tumor histopathology images obtained at the time of initial diagnosis. In all cancer types and all patient groups under study, CHIEF distinguished patients with longer-term survival from those with shorter-term survival. CHIEF outperformed other models by 8 percent. And in patients with more advanced cancers, CHIEF outperformed other AI models by 10 percent. In all, CHIEF’s ability to predict high versus low death risk was tested and confirmed across patient samples from 17 different institutions.

    Extracting novel insights about tumor behavior

    The model identified tell-tale patterns on images related to tumor aggressiveness and patient survival. To visualize these areas of interest, CHIEF generated heat maps on an image. When human pathologists analyzed these AI-derived hot spots, they saw intriguing signals reflecting interactions between cancer cells and surrounding tissues. One such feature was the presence of greater numbers of immune cells in areas of the tumor in longer-term survivors, compared with shorter-term survivors. That finding, Yu noted, makes sense because a greater presence of immune cells may indicate the immune system has been activated to attack the tumor.

    When looking at the tumors of shorter-term survivors, CHIEF identified regions of interest marked by the abnormal size ratios between various cell components, more atypical features on the nuclei of cells, weak connections between cells, and less presence of connective tissue in the area surrounding the tumor. These tumors also had a greater presence of dying cells around them. For example, in breast tumors, CHIEF pinpointed as an area of interest the presence of necrosis — or cell death — inside the tissues. On the flip side, breast cancers with higher survival rates were more likely to have preserved cellular architecture resembling heathy tissues. The visual features and zones of interest related to survival varied by cancer type, the team noted.

    Next steps

    The researchers said they plan to refine CHIEF’s performance and augment its capabilities by:

    • Conducting additional training on images of tissues from rare diseases and non-cancerous conditions
    • Including samples from pre-malignant tissues before cells become fully cancerous
    • Exposing the model to more molecular data to enhance its ability to identify cancers with different levels of aggressiveness
    • Training the model to also predict the benefits and adverse effects of novel cancer treatments in addition to standard treatments
  • DNA tech offers both data storage and computing functions

    DNA tech offers both data storage and computing functions

    Researchers from North Carolina State University and Johns Hopkins University have demonstrated a technology capable of a suite of data storage and computing functions — repeatedly storing, retrieving, computing, erasing or rewriting data — that uses DNA rather than conventional electronics. Previous DNA data storage and computing technologies could complete some but not all of these tasks.

    “In conventional computing technologies, we take for granted that the ways data are stored and the way data are processed are compatible with each other,” says project leader Albert Keung, co-corresponding author of a paper on the work. “But in reality, data storage and data processing are done in separate parts of the computer, and modern computers are a network of complex technologies.” Keung is an associate professor of chemical and biomolecular engineering and a Goodnight Distinguished Scholar at NC State.

    “DNA computing has been grappling with the challenge of how to store, retrieve and compute when the data is being stored in the form of nucleic acids,” Keung says. “For electronic computing, the fact that all of a device’s components are compatible is one reason those technologies are attractive. But, to date, it’s been thought that while DNA data storage may be useful for long-term data storage, it would be difficult or impossible to develop a DNA technology that encompassed the full range of operations found in traditional electronic devices: storing and moving data; the ability to read, erase, rewrite, reload or compute specific data files; and doing all of these things in programmable and repeatable ways.

    “We’ve demonstrated that these DNA-based technologies are viable, because we’ve made one.”

    The new technology is made possible by recent techniques that have enabled the creation of soft polymer materials that have unique morphologies.

    “Specifically, we have created polymer structures that we call dendricolloids — they start at the microscale, but branch off from each other in a hierarchical way to create a network of nanoscale fibers,” says Orlin Velev, co-corresponding author and the S. Frank and Doris Culberson Distinguished Professor of Chemical and Biomolecular Engineering at NC State. “This morphology creates a structure with a high surface area, which allows us to deposit DNA among the nanofibrils without sacrificing the data density that makes DNA attractive for data storage in the first place.”

    “You could put a thousand laptops’ worth of data into DNA-based storage that’s the same size as a pencil eraser,” Keung says.

    “The ability to distinguish DNA information from the nanofibers it’s stored on allows us to perform many of the same functions you can do with electronic devices,” says Kevin Lin, first author of the paper and a former Ph.D. student at NC State. “We can copy DNA information directly from the material’s surface without harming the DNA. We can also erase targeted pieces of DNA and then rewrite to the same surface, like deleting and rewriting information stored on the hard drive. It essentially allows us to conduct the full range of DNA data storage and computing functions. In addition, we found that when we deposit DNA on the dendricolloid material, the material helps to preserve the DNA.”

    “You could say that Keung’s team is providing the equivalent of microcircuits, and the dendricolloidal material that my team creates provides the circuit board,” says Velev. “Our NC State collaborator Adriana San Miguel helped us incorporate the materials into microfluidic channels that direct the flow of nucleic acids and reagents, allowing us to move data and initiate computing commands. Winston Timp’s lab at Johns Hopkins contributed their expertise on nanopore sequencing, which helps us directly read the data in RNA after copying it from DNA on the material’s surface. And James Tuck’s lab — also here at NC State — has developed algorithms that allow us to convert data into nucleic acid sequences and vice versa while controlling for potential errors.”

    The researchers have demonstrated that the new data storage and computing technology — which they call a “primordial DNA store and compute engine” — is capable of solving simple sudoku and chess problems. And testing suggests that it could store data securely for thousands of years in commercially available spaces without degrading the information-storing DNA.

    “What’s more, the dendrocolloidal host material itself is relatively inexpensive and easy to fabricate,” Velev says.

    “There’s a lot of excitement about molecular data storage and computation, but there have been significant questions about how practical the field may be,” says Keung. “We looked back at the history of computing and how the creation of ENIAC inspired the field. We wanted to develop something that would inspire the field of molecular computing. And we hope what we’ve done here is a step in that direction.”

    The paper, “A Primordial DNA Store and Compute Engine,” will be published Aug. 22 in the journal Nature Nanotechnology. The paper was co-authored by Kevin Volkel and Andrew Clark, former Ph.D. students at NC State; Cyrus Cao and Rachel Polak, Ph.D. students at NC State; Adriana San Miguel, an associate professor of chemical and biomolecular engineering at NC State; James Tuck, a professor of electrical and computer engineering at NC State; Winston Timp, an associate professor of biomedical engineering at Johns Hopkins University; and Paul Hook, a postdoctoral researcher at Johns Hopkins.

    The work was done with support from the National Science Foundation under grants 2027655 and 1901324.

    Keung and Tuck are co-founders of DNAli Data Technologies, so have potential interest in translating and commercializing DNA-based information systems. Keung, Volkel, Tuck, and Lin are inventors on patent application WO 2020/096679, which has been licensed to DNAli Data Technologies and from which some of this work is derived.

  • Using AI to find the polymers of the future

    Using AI to find the polymers of the future

    Nylon, Teflon, Kevlar. These are just a few familiar polymers — large-molecule chemical compounds — that have changed the world. From Teflon-coated frying pans to 3D printing, polymers are vital to creating the systems that make the world function better.

    Finding the next groundbreaking polymer is always a challenge, but now Georgia Tech researchers are using artificial intelligence (AI) to shape and transform the future of the field. Rampi Ramprasad’s group develops and adapts AI algorithms to accelerate materials discovery.

    This summer, two papers published in the Nature family of journals highlight the significant advancements and success stories emerging from years of AI-driven polymer informatics research. The first, featured in Nature Reviews Materials, showcases recent breakthroughs in polymer design across critical and contemporary application domains: energy storage, filtration technologies, and recyclable plastics. The second, published in Nature Communications, focuses on the use of AI algorithms to discover a subclass of polymers for electrostatic energy storage, with the designed materials undergoing successful laboratory synthesis and testing.

    “In the early days of AI in materials science, propelled by the White House’s Materials Genome Initiative over a decade ago, research in this field was largely curiosity-driven,” said Ramprasad, a professor in the School of Materials Science and Engineering. “Only in recent years have we begun to see tangible, real-world success stories in AI-driven accelerated polymer discovery. These successes are now inspiring significant transformations in the industrial materials R&D landscape. That’s what makes this review so significant and timely.”

    AI Opportunities

    Ramprasad’s team has developed groundbreaking algorithms that can instantly predict polymer properties and formulations before they are physically created. The process begins by defining application-specific target property or performance criteria. Machine learning (ML) models train on existing material-property data to predict these desired outcomes. Additionally, the team can generate new polymers, whose properties are forecasted with ML models. The top candidates that meet the target property criteria are then selected for real-world validation through laboratory synthesis and testing. The results from these new experiments are integrated with the original data, further refining the predictive models in a continuous, iterative process.

    While AI can accelerate the discovery of new polymers, it also presents unique challenges. The accuracy of AI predictions depends on the availability of rich, diverse, extensive initial data sets, making quality data paramount. Additionally, designing algorithms capable of generating chemically realistic and synthesizable polymers is a complex task.

    The real challenge begins after the algorithms make their predictions: proving that the designed materials can be made in the lab and function as expected and then demonstrating their scalability beyond the lab for real-world use. Ramprasad’s group designs these materials, while their fabrication, processing, and testing are carried out by collaborators at various institutions, including Georgia Tech. Professor Ryan Lively from the School of Chemical and Biomolecular Engineering frequently collaborates with Ramprasad’s group and is a co-author of the paper published in Nature Reviews Materials.

    “In our day-to-day research, we extensively use the machine learning models Rampi’s team has developed,” Lively said. “These tools accelerate our work and allow us to rapidly explore new ideas. This embodies the promise of ML and AI because we can make model-guided decisions before we commit time and resources to explore the concepts in the laboratory.”

    Using AI, Ramprasad’s team and their collaborators have made significant advancements in diverse fields, including energy storage, filtration technologies, additive manufacturing, and recyclable materials.

    Polymer Progress

    One notable success, described in the Nature Communications paper, involves the design of new polymers for capacitors, which store electrostatic energy. These devices are vital components in electric and hybrid vehicles, among other applications. Ramprasad’s group worked with researchers from the University of Connecticut.

    Current capacitor polymers offer either high energy density or thermal stability, but not both. By leveraging AI tools, the researchers determined that insulating materials made from polynorbornene and polyimide polymers can simultaneously achieve high energy density and high thermal stability. The polymers can be further enhanced to function in demanding environments, such as aerospace applications, while maintaining environmental sustainability.

    “The new class of polymers with high energy density and high thermal stability is one of the most concrete examples of how AI can guide materials discovery,” said Ramprasad. “It is also the result of years of multidisciplinary collaborative work with Greg Sotzing and Yang Cao at the University of Connecticut and sustained sponsorship by the Office of Naval Research.”

    Industry Potential

    The potential for real-world translation of AI-assisted materials development is underscored by industry participation in the Nature Reviews Materials article. Co-authors of this paper also include scientists from Toyota Research Institute and General Electric. To further accelerate the adoption of AI-driven materials development in industry, Ramprasad co-founded Matmerize Inc., a software startup company recently spun out of Georgia Tech. Their cloud-based polymer informatics software is already being used by companies across various sectors, including energy, electronics, consumer products, chemical processing, and sustainable materials.

    “Matmerize has transformed our research into a robust, versatile, and industry-ready solution, enabling users to design materials virtually with enhanced efficiency and reduced cost,” Ramprasad said. “What began as a curiosity has gained significant momentum, and we are entering an exciting new era of materials by design.”

  • Researchers develop AI model that predicts the accuracy of protein–DNA binding

    Researchers develop AI model that predicts the accuracy of protein–DNA binding

    A new artificial intelligence model developed by USC researchers and published in Nature Methods can predict how different proteins may bind to DNA with accuracy across different types of protein, a technological advance that promises to reduce the time required to develop new drugs and other medical treatments.

    The tool, called Deep Predictor of Binding Specificity (DeepPBS), is a geometric deep learning model designed to predict protein-DNA binding specificity from protein-DNA complex structures. DeepPBS allows scientists and researchers to input the data structure of a protein-DNA complex into an online computational tool.

    “Structures of protein-DNA complexes contain proteins that are usually bound to a single DNA sequence. For understanding gene regulation, it is important to have access to the binding specificity of a protein to any DNA sequence or region of the genome,” said Remo Rohs, professor and founding chair in the department of Quantitative and Computational Biology at the USC Dornsife College of Letters, Arts and Sciences. “DeepPBS is an AI tool that replaces the need for high-throughput sequencing or structural biology experiments to reveal protein-DNA binding specificity.”

    AI analyzes, predicts protein–DNA structures

    DeepPBS employs a geometric deep learning model, a type of machine-learning approach that analyzes data using geometric structures. The AI tool was designed to capture the chemical properties and geometric contexts of protein-DNA to predict binding specificity.

    Using this data, DeepPBS produces spatial graphs that illustrate protein structure and the relationship between protein and DNA representations. DeepPBS can also predict binding specificity across various protein families, unlike many existing methods that are limited to one family of proteins.

    “It is important for researchers to have a method available that works universally for all proteins and is not restricted to a well-studied protein family. This approach allows us also to design new proteins,” Rohs said.

    Major advance in protein-structure prediction

    The field of protein-structure prediction has advanced rapidly since the advent of DeepMind’s AlphaFold, which can predict protein structure from sequence. These tools have led to an increase in structural data available to scientists and researchers for analysis. DeepPBS works in conjunction with structure prediction methods for predicting specificity for proteins without available experimental structures.

    Rohs said the applications of DeepPBS are numerous. This new research method may lead to accelerating the design of new drugs and treatments for specific mutations in cancer cells, as well as lead to new discoveries in synthetic biology and applications in RNA research.

    About the study: In addition to Rohs, other study authors include Raktim Mitra of USC; Jinsen Li of USC; Jared Sagendorf of University of California, San Francisco; Yibei Jiang of USC; Ari Cohen of USC; and Tsu-Pei Chiu of USC; as well as Cameron Glasscock of the University of Washington.

    This research was primarily supported by NIH grant R35GM130376.

  • Researchers outline promises, challenges of understanding AI for biological discovery

    Researchers outline promises, challenges of understanding AI for biological discovery

    Machine learning is a powerful tool in computational biology, enabling the analysis of a wide range of biomedical data such as genomic sequences and biological imaging. But when researchers use machine learning in computational biology, understanding model behavior remains crucial for uncovering the underlying biological mechanisms in health and disease.

    In a recent article in Nature Methods, researchers at Carnegie Mellon University’s School of Computer Science propose guidelines that outline pitfalls and opportunities for using interpretable machine learning methods to tackle computational biology problems. The Perspectives article, “Applying Interpretable Machine Learning in Computational Biology — Pitfalls, Recommendations and Opportunities for New Developments,” is featured in the journal’s August special issue on AI.

    “Interpretable machine learning has generated significant excitement as machine learning and artificial intelligence tools are being applied to increasingly important problems,” said Ameet Talwalkar, an associate professor in CMU’s Machine Learning Department (MLD). “As these models grow in complexity, there is great promise not only in developing highly predictive models but also in creating tools that help end users understand how and why these models make certain predictions. However, it is crucial to acknowledge that interpretable machine learning has yet to deliver turnkey solutions to this interpretability problem.”

    The paper is a collaboration between doctoral students Valerie Chen in MLD and Muyu (Wendy) Yang in the Ray and Stephanie Lane Computational Biology Department. Chen’s earlier work critiquing the interpretable machine learning community’s lack of grounding in downstream use cases inspired the article, and the idea was developed through discussions with Yang and Jian Ma, the Ray and Stephanie Lane Professor of Computational Biology.

    “Our collaboration began with a deep dive into computational biology papers to survey the application of interpretable machine learning methods,” Yang said. “We noticed that many applications used these methods in a somewhat ad hoc manner. Our goal with this paper was to provide guidelines for more robust and consistent use of interpretable machine learning methods in computational biology.”

    One major pitfall the paper addresses is the reliance on a single interpretable machine learning method. Instead, the researchers recommend using multiple interpretable machine learning methods with diverse sets of hyperparameters and comparing their results to obtain a more comprehensive understanding of the model behavior and its underlying interpretations.

    “While some machine learning models seem to work surprisingly well, we often do not fully understand why,” Ma said. “In scientific domains like biomedicine, understanding why models work is crucial for discovering fundamental biological mechanisms.”

    The paper also warns against cherry-picking results when evaluating interpretable machine learning methods, as this can lead to incomplete or biased interpretations of scientific findings.

    Chen emphasized that the guidelines may have broader implications for a wider audience of researchers interested in applying interpretable machine-learning methods to their work.

    “We hope that machine learning researchers developing new interpretable machine learning methods and tools — particularly those working on explaining large language models — will carefully consider the human-centric aspects of interpretable machine learning,” Chen said. “This includes understanding who their target user is and how the method will be used and evaluated.”

    While understanding model behavior remains crucially important for scientific discovery and a fundamentally unsolved machine learning problem, the authors hope these challenges spur further interdisciplinary collaborations to facilitate the broader use of AI for scientific impact.

  • Cracking the code of life: new AI model learns DNA’s hidden language

    Cracking the code of life: new AI model learns DNA’s hidden language

    DNA contains foundational information needed to sustain life. Understanding how this information is stored and organized has been one of the greatest scientific challenges of the last century. With GROVER, a new large language model trained on human DNA, researchers could now attempt to decode the complex information hidden in our genome. Developed by a team at the Biotechnology Center (BIOTEC) of Dresden University of Technology, GROVER treats human DNA as a text, learning its rules and context to draw functional information about the DNA sequences. This new tool, published in Nature Machine Intelligence, has the potential to transform genomics and accelerate personalized medicine.

    Since the discovery of the double helix, scientists have sought to understand the information encoded in DNA. 70 years later, it is clear that the information hidden in the DNA is multilayered. Only 1-2 % of the genome consists of genes, the sequences that code for proteins.

    “DNA has many functions beyond coding for proteins. Some sequences regulate genes, others serve structural purposes, most sequences serve multiple functions at once. Currently, we don’t understand the meaning of most of the DNA. When it comes to understanding the non-coding regions of the DNA, it seems that we have only started to scratch the surface. This is where AI and large language models can help,” says Dr. Anna Poetsch, research group leader at the BIOTEC.

    DNA as a Language

    Large language models, like GPT, have transformed our understanding of language. Trained exclusively on text, the large language models developed the ability to use the language in many contexts.

    “DNA is the code of life. Why not treat it like a language?” says Dr. Poetsch. The Poetsch team trained a large language model on a reference human genome. The resulting tool named GROVER, or “Genome Rules Obtained via Extracted Representations,” can be used to extract biological meaning from the DNA.

    “GROVER learned the rules of DNA. In terms of language, we are talking about grammar, syntax, and semantics. For DNA this means learning the rules governing the sequences, the order of the nucleotides and sequences, and the meaning of the sequences. Like GPT models learning human languages, GROVER has basically learned how to ‘speak’ DNA,” explains Dr. Melissa Sanabria, the researcher behind the project.

    The team showed that GROVER can not only accurately predict the following DNA sequences but can also be used to extract contextual information that has biological meaning, e.g., identify gene promoters or protein binding sites on DNA. GROVER also learns processes that are generally considered to be “epigenetic,” i.e., regulatory processes that happen on top of the DNA rather than being encoded.

    “It is fascinating that by training GROVER with only the DNA sequence, without any annotations of functions, we are actually able to extract information on biological function. To us, it shows that the function, including some of the epigenetic information, is also encoded in the sequence,” says Dr. Sanabria.

    The DNA Dictionary

    “DNA resembles language. It has four letters that build sequences and the sequences carry a meaning. However, unlike a language, DNA has no defined words,” says Dr. Poetsch. DNA consists of four letters (A, T, G, and C) and genes, but there are no predefined sequences of different lengths that combine to build genes or other meaningful sequences.

    To train GROVER, the team had to first create a DNA dictionary. They used a trick from compression algorithms. “This step is crucial and sets our DNA language model apart from the previous attempts,” says Dr. Poetsch.

    “We analyzed the whole genome and looked for combinations of letters that occur most often. We started with two letters and went over the DNA, again and again, to build it up to the most common multi-letter combinations. In this way, in about 600 cycles, we have fragmented the DNA into ‘words’ that let GROVER perform the best when it comes to predicting the next sequence,” explains Dr. Sanabria.

    The Promise of AI in Genomics

    GROVER promises to unlock the different layers of genetic code. DNA holds key information on what makes us human, our disease predispositions, and our responses to treatments.

    “We believe that understanding the rules of DNA through a language model is going to help us uncover the depths of biological meaning hidden in the DNA, advancing both genomics and personalized medicine,” says Dr. Poetsch.