A team of Australia’s leading health researchers has developed a new ‘dictionary’ to better predict outcomes for people who have experienced a moderate-severe traumatic brain injury (TBI).
The Australian Traumatic Brain Injury Initiative (AUS-TBI) is a consortium of academics, researchers and healthcare professionals from institutions across the country.
Working together to understand the factors which could be used to predict outcome following TBI in a study supported by the Australian Medical Research Future Fund Mission for TBI, the team examined factors related to social support, health, clinical care, biological markers, acute interventions, and longer-term outcomes.
Published across eight articles in a special edition of the Journal of Neurotrauma, this work culminated in a Single Data Dictionary: a master list of information that if collected would give the best chance at understanding how a person will recover from their brain injury.
Consortium lead and Curtin University Deputy Vice-Chancellor, Research Professor Melinda Fitzgerald said predicting each individual’s outcome following a TBI was critical — but also very difficult.
“The rate and degree of recovery after moderate-severe TBI varies greatly, due in part to the complex and diverse nature of these injuries,” Professor Fitzgerald said.
“Despite decades of empirical research, prediction of outcomes after TBI for individual patients remains imprecise: we have only partial understanding of what it is about the person, their injury, their environment, or their care that moderates and/or determines the multiple outcomes that contribute to their quality of life.

“At present, there is no indicator or group of indicators that can sufficiently predict treatment outcome or responsiveness to allow for personalised acute care and rehabilitation for patients with TBI.”
To create the Single Data Dictionary, the AUS-TBI team examined thousands of published research articles reporting factors which may impact a person’s recovery, such as their medical history, social contexts, biological markers, level of personal support, effectiveness of treatment and more.
The team also consulted extensively with people who have lived experience of TBIs, including an Aboriginal and Torres Strait Islander Advisory Group, to develop the Single Data Dictionary of predictive markers alongside guidelines on how to collect them in a coordinated, culturally sensitive national approach.
Co-author Dr Sarah Hellewell, from Curtin’s Faculty of Health Sciences, Curtin Health Innovation Research Institute and the Perron Institute, said providing personalised care to ensure the best outcomes for patients was critical, as TBIs are often catastrophic and have lifelong impact on patients, their families, workplaces, the criminal justice system and society as a whole.
“Severe TBI has a mortality rate of 30-40 per cent, less than 50 per cent of patients achieve long-term independence and new injuries add $2 billion in lifetime direct costs to the Australian healthcare system each year,” Dr Hellewell said.
“Between 2006 and 2015, there was no change in survival or functional outcomes following TBI, proving the need for better, targeted management strategies to reduce mortality and improve quality of life for these individuals and reduce negative impacts on families and society.”
Organisations also involved in the consortium include the Perron Institute, Telethon Kids Institute, Monash University, Griffith University, Deakin University, University of Sydney, University of Queensland, University of Adelaide, University of Melbourne, University of Newcastle, Synapse, Hunter Medical Research Institute, John Hunter Hospital, Brightwater Care Group, The Children’s Hospital at Westmead and Epworth HealthCare.

A study led by linguists from the Department of English, Linguistics and Theatre Studies (ELTS) at the NUS Faculty of Arts and Social Sciences (FASS) has found that early linguistic signs of dementia can be detected through the study of the natural speech of senior Singaporeans.
This groundbreaking study, conducted together with researchers from the NUS Yong Loo Lin School of Medicine (NUS Medicine), compared the natural speech of cognitively healthy persons with those suffering from mild cognitive impairment (MCI) to detect linguistic markers of dementia. It revealed that those with memory-related MCI spoke less and produced fewer, but more abstract, nouns — a speech pattern that is consistent with patients diagnosed with Alzheimer’s disease, a specific type of dementia.
The study’s principal investigator, NUS Department of ELTS Professor Bao Zhiming, noted that Singapore provides a unique environment for this research given the varied use of languages here, with four official languages and a blend of various dialects. He added, “Previous studies had analysed targeted and smaller volumes of language data through word-based fluency tests, structured interviews and picture narrations. Our study has never been done before as it focused on unstructured and spontaneous speech that is easy to collect and analyse.”
Team member, Yeo Boon Khim Mind Science Centre (YBK MSC) Advisory Board Member and NUS Medicine’s psychiatrist Emeritus Professor Kua Ee Heok, said, “There is an urgent need for innovative strategies to combat the rising rate of dementia in Singapore given our rapidly aging population. As research data for this study were taken from participants of a broader YBK MSC research project known as the Community Health Intergenerational (CHI) Study, led by Dr Rathi Mahendran, the findings will ultimately contribute to the CHI study’s goal of identifying at-risk seniors and implementing interventions that can help the elderly age well.”
The study was published in the journal Alzheimer’s & Dementia: Diagnosis, Assessment and Disease Monitoring on 18 April.
Compiling and analysing natural speech data
The team obtained natural speech data from 148 elderly Singaporeans in their 60s and 70s — half of them were cognitively healthy (individuals who have the ability to think clearly, learn and remember) while the other half of the participants had MCI.

Out of the 74 subjects with MCI, 38 had been diagnosed with amnestic MCI (MCI that affects the memory) while 36 had been diagnosed with non-amnestic MCI (MCI that affects thinking skills other than memory). Amnestic MCI carries a higher risk of conversion to Alzheimer’s disease while non-amnestic MCI is linked to a higher risk of conversion to other types of dementia such as Lewy Body Dementia. Overall, an estimated 10 to 20 per cent of people aged 65 or older with MCI go on to develop dementia.
Participants were instructed to speak about any topic in English for 20 minutes with minimal involvement from interviewers and these were recorded with simple digital voice recorders in an ordinary office setting. Topics varied freely and widely, ranging from work and retirement to family life and public affairs.
The recordings yielded 267,310 words which were then transcribed and then tagged as noun or verb using a Part-of-Speech tagger software. The team then calculated the per-minute word counts and concreteness scores of all tagged words.
Early signs of dementia detected in people with amnestic MCI
Findings revealed that participants with amnestic MCI spoke less, produced fewer and more abstract nouns than people with non-amnestic MCI and the healthy controls. Verbs were not affected. A problem with imageability, which is the degree to which a word’s meaning evokes a mental image, was detected in natural everyday speech by people with amnestic MCI.
Dr Luwen Cao, also from the NUS Department of ELTS, said, “Our findings are a significant breakthrough as traditional diagnoses of dementia are done following a battery of neuropsychological and neurological tests. The study of natural speech to detect linguistic signs of early cognitive decline is a reliable, non-invasive and cost-effective tool that could possibly help medical practitioners in the early diagnosis, intervention and management of the progressive disease.”
Moving forward, the team plans to work with the neurologists at the YBK MSC to device language-based intervention strategies to address the language difficulties experienced by people with amnestic MCI.
Prof Bao said, “Ultimately, our research aims to contribute to healthy aging in Singapore. Singapore is aging fast; a quarter of Singaporeans are over 60 years old. By exploring innovative diagnostic tools and intervention strategies, we hope to improve the quality of life for older adults and reduce the burden on healthcare systems. Our work is one step towards ensuring that our aging population enjoy longer, healthier lives.”

A team of researchers led by LEE Soo Ahn and WOO Choong-Wan at the Center for Neuroscience Imaging Research (CNIR) within the Institute for Basic Science (IBS), in collaboration with CHOI Myunghwan at Seoul National University and Tor D. WAGER at Dartmouth College, has revealed how the brain processes emotional information of sustained pain and pleasure.
Using functional Magnetic Resonance Imaging (fMRI), the team recorded brain activity while participants experienced sustained pain and pleasure induced by capsaicin and chocolate fluids. Through sophisticated machine learning techniques, they unraveled the brain activity patterns that encode pleasant or unpleasant emotions and the magnitude of sustained pain and pleasure.
Although pain and pleasure are opposite experiences, they are intricately connected. Previous studies have suggested a set of brain regions that respond to both pain and pleasure. However, most previous studies have been conducted on animals rather than humans, and studies that directly compared the brain representations of pain and pleasure within the same individuals are still lacking.
In this study, the research team conducted an experiment that induced sustained pain and pleasure to participants in the MR scanner, by delivering capsaicin and chocolate fluids. While experiencing sustained pain and pleasure, participants reported moment-by-moment changes in subjective pleasantness and unpleasantness. The participants’ subjective reports of pleasantness and unpleasantness gradually increased and persisted during the capsaicin and chocolate fluids deliveries and decreased after the deliveries ended. By inducing dynamic changes in sustained pain and pleasure, the team aimed to identify the brain regions activated by both experiences.
The research team collected the brain imaging data and moment-by-moment changes in pleasantness or unpleasantness ratings from 58 participants. The team utilized machine learning techniques to analyze the brain data, and they identified a set of brain regions that responded to both sustained pain and pleasure. Based on the brain activity patterns of these common brain regions, the team developed two predictive models to capture (1) the magnitude of affective experiences regardless of how pleasant or unpleasant they are (i.e., ‘affective intensity’) and (2) the magnitude of pleasantness or unpleasantness (i.e., ‘affective valence’).
The researchers found that these models successfully predicted the affective intensity and valence information of sustained pleasure and pain, from both the 58 individuals in the training dataset and 61 new individuals in the independent test dataset. The activity patterns predictive of the affective intensity and valence were spatially distinguishable, and these patterns were connected to distinct functional brain networks. This suggests that the affective intensity and valence information represent multiple aspects of brain mechanisms underlying pain-pleasure interaction.
“While there have been separate lines of studies on pain and pleasure, research comparing the experiences of both pain and pleasure within the same individuals has been rarely conducted,” stated Dr. WOO Choong-Wan, associate director of IBS, who led the study. “The brain activity patterns for affective valence and intensity can contribute to the understanding of how pain and pleasure interact, as well as the brain mechanisms underlying depression commonly observed in chronic pain patients.”
LEE Soo Ahn, a doctoral candidate and the first author of this study, emphasized, “These results demonstrate that pain and pleasure share the same underlying emotional information on pleasantness and unpleasantness,” adding, “We should focus on the fact that affective valence and intensity information can be represented across multiple brain regions.”

The selective bond-breaking powers of enzymes bring new versatility for building nanoparticles with a wide range of technical and medical potential.
Researchers at Hokkaido University have developed a new and more adaptable method for creating nanoparticles of finely controlled size. Their ‘bio-catalytic nanoparticle shaping’ (BNS) procedure, published in the journal Nanoscale Horizons, should greatly assist the production of a variety of nanoparticles for use in technology and medicine.
“One of the most promising applications is for creating assemblies of nanoparticles called quantum dots, which are small enough for their properties to be influenced by subtle quantum mechanical effects,” says Associate Professor Yuta Takano, the leader of the Hokkaido team. Takano and colleagues collaborated on the work with researchers at the University of Melbourne in Australia.
The method uses enzymes to cut specific chemical bonds in molecular assemblies that have been made by linking together small organic (carbon-based) molecules, inorganic materials, or mixed organic and inorganic components. A variety of natural and readily-obtained enzymes can cut the linker sections of the original larger structures into nanoparticles of various desired sizes and shapes.
Varying the linkers and the core components held together by the linkers allows a range of different nanoparticles to be created, each with differing chemical and physical properties and differing sizes.
The researchers made several types of nanoparticles to demonstrate the potential of their technique. One example formed quantum dots whose optical and electronic properties could be useful in technological applications including molecular computation, high density data storage devices, photocatalysis and solar cells.
They also explored nanoparticles with chemical effects that could potentially be used to kill cancer cells or specific disease-causing bacteria. Another feasible medical application is to attach drugs to the nanoparticles, allowing them to achieve targeted drug delivery directly to the specific sites of disease.
“The potential of the bio-catalytic nanoparticle approach is enormous,” concludes Takano, “drawing on the chemical variability and power of naturally available enzymes opens a whole new area of opportunity in nanoparticle design and production.”
Takano has filed a patent application as the inventor of the new technique.
The researchers will now continue their exploration of this exciting new frontier, expanding the possibilities themselves while hopefully also encouraging other teams to pick up on the concept and develop their own ideas. They hope to eventually see bio-catalytic approaches commercialized and being exploited in many areas of research, technology and medicine.

Flinders University researchers have analysed how an antidiabetic treatment could help control the growth of tumours, potentially paving the way for the design of better cancer treatments.
The new study investigated what happens when metformin, a type 2 diabetes medication, is used to treat colorectal cancer cells, in the process demonstrating that it could be exploited to develop new cancer therapies.
Previous epidemiology studies show that taking metformin helps protect diabetes patients from developing some forms of cancer including bowel, or colorectal, cancer.
The Flinders’ researchers sought to understand how taking metformin medication impacts cancer cells and how this could help with future cancer treatments.
“Using the latest techniques, we analysed how metformin helps to stop colorectal cancer cells from growing and multiplying by controlling certain ‘pathways’ inside the cells that help to regulate growth and division,” says lead author Dr Ayla Orang from Flinders University’s College of Medicine and Public Health.
“Importantly, our work has pinpointed that metformin uses small pieces of RNA (called microRNAs) to act as a ‘circuit breaker’ and turn off certain genes that are involved in cell growth and division, so it is possible that our findings could eventually be used to develop a new targeted cancer therapy.
“In particular, we found that metformin increases the levels of certain microRNAs, like miR-2110 and miR-132-3p, which then target specific genes and slow down the growth and progression of tumours.

“With this information we may be able to develop RNA-based therapies — new treatments for cancer that target RNA molecules (like microRNAs),” she says.
The research, Restricting Colorectal Cancer Cell Metabolism with Metformin: An Integrated Transcriptomics Study, used advanced techniques to study microRNAs, and the entire set of genes being expressed in the colon cancer cells, to help understand how metformin affects the cells.
Metformin increased the levels of certain microRNAs (miR-2110 and miR-132-3p) that target a specific gene (PIK3R3).
This process helps to slow down the growth of cancer cells and stop them from multiplying too quickly. Another gene (STMN1) was also targeted by different microRNAs, which led to slower cell growth and a delayed cell cycle.
Senior authors of the study, Associate Professor Michael Michael and Professor Janni Petersen say the results are a step forward in our understanding of the way metformin disrupts cancer cell growth and how they could be used to fight cancer.
“Our research provides new insights into the molecular mechanisms of how metformin works, and how we might be able to target genes responsible for turning normal cells cancerous,” says Associate Professor Michael.
“This is important because it shows the potential of metformin as a preventive agent for reducing the growth of cancer in the bowel, and the emergence of RNA therapeutics as a promising new avenue for exploring the clinical efficacy of these findings.
“We need to further investigate the potential therapeutic benefits of targeting specific miRNAs or pathways using RNA-based therapies for the treatment of cancer.
Having used metformin to unravel metabolism in cancer cells, the next stage of research is focusing on specific cell pathways, which should lead to animal studies and then human clinical trials.”

Chronic lung diseases are often accelerated and exacerbated by polymicrobial infections. An international study team led by MedUni Vienna has identified two types of these so-called dysbioses in cystic fibrosis. They display distinct ecology and are also likely to respond differently to treatment. The study was published in the journal Nature Communications.
Chronic lung diseases such as COPD, asthma or cystic fibrosis affect many people worldwide. In 2019, 454.6 million cases were registered worldwide. These diseases display a progressive loss of lung function and are associated with a high mortality rate. Polymicrobial infections of the respiratory tract, in which bacterial communities persist in the lungs for long time, constitute a major risk factor and are difficult to treat. These infections are often associated with recurrent, acute worsening of symptoms (exacerbations, “PEx”), which have a negative impact on the course of the disease.
A recent study led by Stefanie Widder from MedUni Vienna in collaboration with colleagues led by John J. LiPuma from the University of Michigan Medical School Ann Arbor has focused on the characterization of disease-associated bacterial communities (“dysbioses”) in subjects with cystic fibrosis and investigated their ecological networks. The aim was to develop hypotheses that enable the development of more precise treatment strategies for people with chronic lung diseases.
Two different types of dysbiosis
To this end, sputum samples (mucus expectorated from the lung) were collected from people with cystic fibrosis over an extended period of time, sequenced and then analyzed using computational models by Stefanie Widder (Department of Medicine I, Division of Infection Biology, MedUni Vienna). Two antagonistic types of dysbiosis were discovered, which differ fundamentally in their organization: they either form hierarchical or stochastic networks. The structural differences of the microbiota have far-reaching consequences: Based on the sequencing data, it was shown that important pathogens such as Pseudomonas aeruginosa or Staphylococcus aureus only served as drivers of infection if located on top of the microbial hierarchy. In less structured microbiota, they showed more random dynamics, suggesting that they might be less decisive for the course of the infection.
Computer model predicts different responses to treatments
Moreover, the two types of dysbiosis are likely to respond differently to treatments. A simplified computer model that simulated the effect of antimicrobial drugs on pathogens predicted better efficacy with hierarchically organized microbiota. Both aspects are very relevant for patients and their medical care teams: “Our study reveals a data-based, causal relationship between PEx, microbial ecology and treatment success,” explains study leader Stefanie Widder, “the findings form an important basis for further translational research into personalized management of dysbiosis, both in cystic fibrosis and in other obstructive pulmonary diseases.”

A computer program based on data from nearly a half-million tissue images and powered by artificial intelligence can accurately diagnose cases of adenocarcinoma, the most common form of lung cancer, a new study shows.
Researchers at NYU Langone Health’s Perlmutter Cancer Center and the University of Glasgow developed and tested the program. They say that because it incorporates structural features of tumors from 452 adenocarcinoma patients, who are among the more than 11,000 patients in the United States National Cancer Institute’s Cancer Genome Atlas, the program offers an unbiased, detailed, and reliable second opinion for patients and oncologists about the presence of the cancer and the likelihood and timing of its return (prognosis).
The research team also points out that the program is independent and “self-taught,” meaning that it determined on its own which structural features were statistically most significant to gauging the severity of disease and had the greatest impact on tumor recurrence.
Publishing in the journal Nature Communications online June 11, the study program, also called an algorithm, or specifically, histomorphological phenotype learning (HPL), was found to accurately distinguish between similar lung cancers, adenocarcinoma and squamous cell cancers, 99% of the time. The HPL program was also found to be 72% accurate at predicting the likelihood and timing of cancer’s return after therapy, bettering the 64% accuracy in the predictions made by pathologists who directly examined the same patients’ tumor images, researchers say.
“Our new histomorphological phenotype learning program has the potential to offer cancer specialists and their patients a quick and unbiased diagnostic tool for lung adenocarcinoma that, once further testing is complete, can also be used to help validate and even guide their treatment decisions,” said study lead investigator Nicolas Coudray, PhD, a bioinformatics programmer at NYU Grossman School of Medicine and Perlmutter Cancer Center.
“Patients, physicians, and researchers know they can rely on this predictive modeling because it is self-taught, provides explainable decisions, and is based only on the knowledge drawn specifically from each patient’s tissue, including such features as its proportion of dying cells, tumor-fighting immune cells, and how densely packed the tumor cells are, among other features,” said Coudray.
“Lung tissue samples can now be analyzed in minutes by our computer program to provide fairly accurate predictions of whether their cancer will return, predictions that are better than current standards of care for making a prognosis in lung adenocarcinoma,” said study co-senior investigator Aristotelis Tsirigos, PhD. Tsirigos is a professor in the Departments of Pathology and Medicine at NYU Grossman School of Medicine and Perlmutter Cancer Center, where he also serves as co-director of precision medicine and director of its Applied Bioinformatics Laboratories.

Tsirigos says that thanks to such tools and other advances in the lung cancer biology, pathologists will be examining tissue scans on their computer screens, and less and less on microscopes, and then using their AI program to analyze the image and produce its own image of the scan. The new image, or “landscape,” they add, will offer a detailed breakdown of the tissue’s content. It might note, for example, that there is 5% necrosis and 10% tumor infiltration and what that means in terms of survival. That reading may statistically equate to an 80% chance of remaining cancer-free for two years or more, based on information from all the patient data in the program.
To develop the HPL program, the researchers first analyzed lung adenocarcinoma tissue slides from the Cancer Genome Atlas. Adenocarcinoma was chosen for the test model because the disease is known for characteristic features. As an example, they note that its tumor cells tend to group in so-called acinar, or saclike patterns and spread predictably along the surface lining of lung cells.
From their analysis of the slides, whose visual images were digitally scanned and broken into 432,231 small quadrants or tiles, researchers found 46 key characteristics, what they term histomorphological phenotype clusters, from both normal and diseased tissue, a subset of which were statistically linked to either cancer’s early return or to long-term survival. The findings were then confirmed by further and separate testing on tissue images from 276 men and women who were treated for adenocarcinoma at NYU Langone from 2006 to 2021.
Researchers say their goal is to use the HPL algorithm to assign to each patient a score between 0 and 1 that reflects their statistical chance of survival and tumor recurrence for up to five years. Because the program is self-learning, they stress HPL will become increasingly more accurate as more data is added over time. To build public trust, researchers have posted their programming code online and have plans to make the new HPL tool freely available upon completion of further testing.
Characteristics linked to tumors recurring included high tile percentages of dead cancer cells and tumor-fighting immune cells called lymphocytes, and the dense clustering of tumor cells in the outer linings of the lungs. Features tied to increased likelihood for survival were high percentages of unchanged or preserved lung sac tissue, and lack of or mild presence of inflammatory cells.
Tsirigos says the team next plans to look at developing HPL-like programs for other cancers, such as breast, ovarian, and colorectal, that are similarly based on distinctive and key morphological features and additional molecular data. The team also has plans to expand and improve the accuracy of the current adenocarcinoma HPL program by including other data from hospital electronic health records about other illnesses and diseases, or even income and home ZIP code.
Funding support for the new study was provided by National Institutes of Health grant P30CA016087, United Kingdom Research Council grants Ep/R018634/1 and BB/V016067/1, and European Union Horizon 2020 grant no. 101016851.
Besides Tsirigos and Coudray, other NYU Langone researchers involved in this study are Anna Yeaton, Bojing Liu, Hortense Le, Luis Chiriboga, Afreen Karimkhan, Navneet Natula, Christopher Park, Harvey Pass, and Andre Moreira. Study co-lead investigator Adalberto Claudio Quiros, study co-investigators Xinyu Yang and John Le Quesne, and study co-senior investigator Ke Yuan are all at the University of Glasgow, UK. Study co-investigator David Moore is at the University College London, UK.

Expression of Gstt1 enables the cells to change the surrounding environment to support their growth. The findings could lead to new strategies to treat metastatic cancer and would be particularly impactful for patients with pancreatic cancer.
Metastatic cancer cells, which cause 90% of cancer-related deaths, must overcome numerous hurdles to spread from a primary tumor through the bloodstream and re-establish themselves in different tissues.
A new study led by investigators from the Mass General Cancer Center has identified a gene whose expression confers a growth advantage to these cells.
Mechanistically, the gene’s expression allows metastatic cancer cells to cause changes to their surrounding environment so that they can grow in new locations in the body. The findings are published in Nature Cell Biology.
“Our results point to potentially novel therapeutic avenues to specifically target metastatic cancer,” said senior author Raul Mostoslavsky, MD, PhD who is the scientific director of the Krantz Family Center for Cancer Research at the Mass General Cancer Center.
Mostoslavsky and colleagues first compared gene expression patterns in primary versus metastatic tumors in mice with pancreatic cancer or breast cancer. After identifying various genes whose expression increased in metastatic tumor cells, the researchers silenced each gene individually.
In these experiments, silencing the Gstt1 gene had no effect on primary tumor cells from mice, but it stripped metastatic cancer cells of their ability to grow and spread. It also blocked cell growth in two metastatic-derived human pancreatic cancer cell lines.
Gstt1 encodes an enzyme that is a member of a superfamily of proteins involved with protecting cells from toxins, among other functions. Mechanistic studies indicated that the Gstt1 enzyme causes metastatic cancer cells to modify and secrete a protein called fibronectin, which is important for helping cells to attach themselves to the extracellular matrix, a large network of proteins and other molecules that surround, support, and give structure to cells and tissues in the body.
“Gstt1 alters the matrix surrounding the metastatic cells so they can grow in these foreign niches,” said Mostoslavsky. “Our results could lead to new strategies for the treatment of metastatic disease. This would be especially impactful for pancreatic cancer, in which most patients present with metastases when initially diagnosed.”

Using an advanced artificial intelligence tool, researchers at the Icahn School of Medicine at Mount Sinai have identified rare coding variants in 17 genes that shed light on the molecular basis of coronary artery disease (CAD), the leading cause of morbidity and mortality worldwide.
The discoveries, detailed in the June 11 online issue of Nature Genetics, reveal genetic factors impacting heart disease that open new avenues for targeted treatments and personalized approaches to cardiovascular care.
The investigators used an in silico, or computer-derived, score for coronary artery disease (ISCAD) that holistically represents CAD, as described in a previous paper by the team in The Lancet. The ISCAD score incorporates hundreds of different clinical features from the electronic health record, including vital signs, laboratory test results, medications, symptoms, and diagnoses. To build the score, they trained machine learning models on the electronic health records of 604,914 individuals across the UK Biobank, All of Us Research Program, and BioMe Biobank in this comprehensive meta-analysis.
The score was then tested for association with rare and ultra-rare coding variants found in the exome sequences of these individuals. In addition, the research team conducted further investigation into the discovered genes to study their roles in causal CAD risk factors, clinical manifestations of CAD, and their connections with CAD status in traditional large-scale genome-wide association studies, among other factors.
“Our findings help us understand how these 17 genes are involved in coronary artery disease. Some of these genes are already known to influence heart disease development, while others have never been linked to it before,” says Ron Do, PhD, senior study author and the Charles Bronfman Professor in Personalized Medicine at Icahn Mount Sinai. “Our study shows how machine learning tools can uncover genetic insights that traditional methods might miss when comparing cases and controls. This could lead to new ways to identify biological mechanisms of heart disease or gene targets for treatment.”
Because they occur in only a small percentage of individuals, rare coding variants may have a significant impact on disease risk or susceptibility when present. Therefore, studying these variants is essential to understanding the genetic basis of diseases and can inform therapeutic targets.
The study was driven by the challenges faced, over the last decade, in identifying rare coding variants associated with CAD using traditional methods relying on diagnosed cases and controls. Diagnostic codes’ limitations in capturing the complexity of CAD prompted the researchers to explore new avenues of investigation.
“Our previous Lancet paper showed that a machine learning model trained with electronic health records can generate an in silico score for coronary artery disease, capturing disease across its spectrum,” says lead author Ben Omega Petrazzini, BS, Associate Bioinformatician in Dr. Do’s lab at Icahn Mount Sinai. “Based on these findings, we hypothesized that the in-silico score for CAD could reveal novel rare coding variants related to CAD by offering a more holistic view of the disease.”
Next, the investigators plan to further investigate the role of the identified genes in CAD biology and explore potential applications of machine learning in the genetic study of other complex diseases, as part of their ongoing efforts to advance understanding of disease mechanisms, discover new treatments, and improve patient outcomes.

10 minutes agoBy