Machine Learning in Health Care

Groundbreaking study reveals hidden complexity in human genetics

Teresa Donnellan — Mon, 12 Jan 2026 18:32:03 +0000

Groundbreaking study reveals hidden complexity in human genetics Teresa Donnellan Mon, 01/12/2026 - 13:32

In This Story

Sometimes, in genetics, two wrongs do make a right. A research team recently showed that two harmful genetic variants, when occurring together in a gene, can restore function—proving a decades-old hypothesis originally proposed by Nobel laureate Francis Crick. Their study, published in the Proceedings of the National Academy of Sciences (PNAS), not only experimentally validated this theory but also introduced a powerful artificial intelligence (AI)-driven approach to genetic interpretation led by �鶹��Ƶ researchers.

The project began when Aimée Dudley, a geneticist at the Pacific Northwest Research Institute (PNRI), approached �鶹��Ƶ Chief AI Officer Amarda Shehu after following her lab’s work on frontier AI models for predicting the functional impact of genetic variation. That conversation sparked a collaboration that married PNRI’s experimental expertise with George �鶹��Ƶ’s computational innovation to discover some surprising ways variant combinations can shape human health.

The problem

Every year one in three Americans is diagnosed with a genetic disorder. Symptoms manifest in infancy for about 70% of individuals. Sadly, 35% die before the age of 5. Advancements in clinical genomics offer hope to better understand and possibly treat these disorders.

“High-throughput genomic screening has been a wonderful feat for humanity,” said Shehu, “but one of its side effects is that it has produced massive amounts of data, outpacing our ability to interpret what that data means for health and disease.”

Research in the Shehu lab has for years focused on building frontier AI models to advance genetic interpretation, but all data available link only isolated, single variants to measured functional activity. Because each person's genome contains billions of base pairs, with about five million variants existing between two individuals’ genomes, looking at one variant at a time rather than combinations of variants could only reveal so much.

“It looked like we had hit a wall,” Shehu said, “that is, until Dr. Dudley contacted my lab more than a year ago.”

On the left, a 3D landscape derived from variant–variant measurements shows distinct functional regions emerging from pairwise interactions. On the right, these regions map onto a multimeric protein structure, where variants in separate spatial zones can be sequestered into different active sites, allowing functional recovery. This visualization captures the structural logic underlying positive epistasis and illustrates how AI-enabled analysis links genetic variation to protein function, a key, groundbreaking result Dudley and Shehu's labs published in the Proceedings of the National Academy of Sciences. Image provided.

The proof

Dudley’s lab was convinced that the key was to account for variant combinations in a gene, also called epistasis. They measured functional effects of variant combinations in the DNA of a key enzyme, argininosuccinate lyase (ASL), a lack of which results in urea cycle disorder, a rare but devastating condition.

The researchers tested thousands of variant combinations that resulted in no enzyme activity when on their own and found that a significant portion of them had high levels of enzyme activity when in combination with each other. In other words, two defective variants, when combined, can recover function.

“This was the most puzzling thing that I could not believe when Dr. Dudley showed it to me. Sometimes in biology, zero plus zero equals 100%,” said Shehu.

Shehu said that Crick, who shared the Nobel Prize in Physiology or Medicine 1962 with James Dewey Watson and Maurice Wilkins for their discoveries concerning the molecular structure of DNA, had hypothesized this could happen.

“Crick had a fancy word for it—variant sequestration,” said Shehu, “But until Dr. Dudley, no one had demonstrated it.”

The progress

Once Dudley’s lab confirmed the phenomenon experimentally, George �鶹��Ƶ researchers turned to AI to see if it could predict similar effects across other genes. Using the ASL data from Dudley’s lab, George �鶹��Ƶ computer science PhD student Anowarul Kabir developed a machine learning model to predict the effects of variant combinations. Then, he applied the model to a structurally similar but evolutionarily distinct protein, fumarase (FH). The algorithm achieved 99.6 percent accuracy in predicting regained function within ASL and 91 percent accuracy in FH.

“The really cool thing about this,” said Shehu, “is that the model learned both sequence and structural patterns and was able to transfer its knowledge to another gene.”

This breakthrough suggests that with experimental data from a few genes, AI can help scale variant effect prediction to a broad set of genes. The PNAS publication estimates that as many as 4% of the genes in the human genome could have the same types of effects seen for ASL and FH.

The paradigm shift

This breakthrough marks a paradigm shift in clinical genomics for precision medicine. By considering variant combinations rather than isolated, single variants, clinicians can deliver faster, more accurate diagnoses and life-saving interventions for families facing rare diseases. They can also prioritize therapeutic treatments based on specific epistatic profiles of patients or clinical trial participants.

“Clinical genomics has been stuck in a rut for decades. We’ve shown that you need to look at combinations of variants to fully understand their impact,” said Shehu. “Our AI model expands coverage from one gene to another, accelerating interpretation and bringing us closer to true precision medicine.”

Topics

DNA

Machine Learning

Machine Learning in Health Care

Research

Artificial Intelligence

Determining quality in forensic injury imaging - �鶹��Ƶ secures NIH AIM-AHEAD funding to advance equity in AI-driven injury detection

dhawkin — Fri, 22 Nov 2024 18:21:21 +0000

Determining quality in forensic injury imaging - �鶹��Ƶ secures NIH AIM-AHEAD funding to advance equity in AI-driven injury detection dhawkin Fri, 11/22/2024 - 13:21

In This Story

Janusz Wojtusiak, PhD

Katherine Scafide, PhD, RN, FAAN, FAAFS

David Lattanzi

�鶹��Ƶ has been awarded funding from the National Institutes of Health (NIH) through its Artificial Intelligence/Machine Learning Consortium to Advance Health Equity and Researcher Diversity (AIM-AHEAD) program.

This initiative encourages more people from historically underrepresented groups in researching and developing of artificial intelligence (AI) and machine learning (ML) models. AIM-AHEAD aims to leverage the growing volume of data generated through electronic health records (EHR) and other biomedical research to address health disparities and inequities.

Janusz Wojtusiak. Photo by the Office of University Branding

The newly funded project, led by Janusz Wojtusiak, builds upon the Equitable and Accessible Software for Injury Detection (EAS-ID) initiative aimed at building artificial intelligence tools for collecting, assessing and analyzing injury data. The new funding specifically addresses the problem of measuring equity and quality of imaging documentation.

In addition to Wojtusiak, an interdisciplinary research team comprising Katherine Scafide and David Lattanzi is joined by Health Informatics Assistant Professor Eman Elashkar, Research Assistant Professor Jesse Kirkpatrick, who is also the acting director of the Institute for Philosophy and Public Policy, and Amin Nayebi Nodoushan, a postdoctoral researcher at �鶹��Ƶ.

Their research focuses on using AI methods combined with Alternate Light Sources (ALS) to improve bruise detection, addressing visibility issues for individuals with darker skin tones who often encounter challenges in accurately assessing injuries sustained from violence.

Current literature highlights that skin color significantly influences the accuracy of AI-based tools in healthcare. Studies have documented disparities in the performance of medical devices, such as pulse oximeters and smartwatches, which frequently yield inaccurate readings for individuals with darker skin. These discrepancies can lead to delays in critical medical interventions, exacerbating existing health disparities.

The proposed research will focus specifically on bruises, the most common type of soft tissue injury experienced by victims of intimate partner violence (IPV). Statistics indicate that approximately one in three people in the U.S. have experienced IPV, with racial minorities reporting disproportionately higher rates. Survivors with darker skin tones have noted that their bruises are often invisible, resulting in significant delays in seeking necessary medical care.

Katherine Scafide is part of the multidisciplinary team working to advance equity in AI-driven injury detection. Photo by Emma Anderson.

Leveraging the innovative use of ALS, the George �鶹��Ƶ research team has demonstrated marked improvement in bruise visibility across diverse skin tones. The team aims to develop methods that ensure that AI-based tools provide equitable and unbiased detection and characterization of injuries. This will involve creating combined technical-ethical metrics to assess the performance of these tools across different skin tones. Engaging diverse stakeholders, including clinicians, forensic nurses, and community representatives, will be essential in the development process to align with ethical practices in AI.

The research team’s two primary aims are to develop metrics that assess equity in AI tools and apply these metrics to improve bruise detection models. They have already collected a substantial dataset of bruise images taken under various lighting conditions, which will be utilized to enhance the AI models’ performance. The interdisciplinary nature of the research team, comprising informaticians, engineers, clinicians, and ethicists, ensures a comprehensive approach to tackling these complex issues.

In alignment with AIM-AHEAD’s goals, this research initiative promises to contribute significantly to addressing health inequities and enhancing the capabilities of AI in healthcare. By focusing on the specific needs of underrepresented communities, the George �鶹��Ƶ research team is paving the way for more equitable health care solutions, ultimately aiming to improve the accuracy and efficacy of injury assessments across diverse populations.

This project is led by George �鶹��Ƶ’s College of Public Health in collaboration with the College of Engineering and Computing. More information on the project can be found at bruise.gmu.edu.

Topics

Bruise Visibility

Machine Learning

Machine Learning in Health Care

Research

Artificial Intelligence

Symptoms and Movement Patterns During the COVID-19 Pandemic

dhawkin — Mon, 22 Feb 2021 21:55:42 +0000

Symptoms and Movement Patterns During the COVID-19 Pandemic dhawkin Mon, 02/22/2021 - 16:55

In This Story

Janusz Wojtusiak, PhD

How can we better understand how people move during the pandemic and how they spread COVID-19? Dr. Janusz Wojtusiak, associate professor of health informatics and director of the Machine Learning Inference Lab is leading one of the first individual-level studies on social distancing.

Participants use the �鶹��Ƶ COVID Health Check^TM to record symptoms of possible COVID-19 infection and GPS and WiFi data to provide information on how they move during the pandemic. This allows the researchers to model and predict movements during the pandemic and in conjunction with any reported possible COVID-19 symptoms. This could help inform effective public health interventions to reduce infection.

Initial findings from the first wave of the study, including data collected through September 2020, were published in the Journal of Healthcare Informatics Research. Wojtusiak and colleagues found that headache was the most frequently reported symptom and headache was always listed as a symptom when any other symptoms were reported. The next most commonly reported symptoms were cough and sore throat.

Movement patterns varied among participants, with some only going out for essential trips while others moved about more. As a group, movement was consistent over the study period, which included a period when Virginia was under a stay-at-home order and when it was not. Participants traveled a total average of 139 miles per week, visiting an average of less than six locations per week. This low average mileage and number of sites visited does suggest that COVID-19-related restrictions affected their movement. However, they also found that even when participants reported symptoms of COVID-19 or contact with others with COVID-19, they did not change their movements as recommended by public health guidance.

This research is possible thanks to the dedication of study participants who share their data to allow for movement modeling. Recruitment of �鶹��Ƶ faculty, staff, and students for the second wave of the study has begun. Learn more and sign up: https://www.mli.gmu.edu/sd/.

�鶹��Ƶ has a very low COVID-19 infection rate, and during the period none of the study participants reported COVID-19 infection, so researchers weren’t able to link COVID-19 positive tests and movement. Future analysis will include data from the 2020 winter so may provide more information on movement after COVID-19 infection. The researchers will also conduct surveys and interviews to provide richer data including reasons for complying or not complying with social distancing.

In a related study supported by the National Cancer Institute, Wojtusiak’s team analyzed individual movements of people on campus. Such micro-scale movements within buildings can be modeled using WiFi data collected each time a mobile phone or laptop is connected to the internet. In over 150 simulated scenarios they were able to reconstruct movements of volunteers within �鶹��Ƶ’s Peterson Hall. This technology is intended to support contact elicitation as part of contact tracing for COVID-19 or other infectious diseases.

Promising preliminary results show that the technology can change how public health officials think about contact tracing. Tests are now being conducted across other locations on Fairfax campus. You can learn more about the project at https://www.mli.gmu.edu/wifi.

Topics

CHHS Research

Machine Learning

Machine Learning in Health Care

coronavirus; covid-19

College of Health and Human Services

College of Health and Human Services Department of Health Administration and Policy

CHHS News

Panagiota Kitsantas, PhD

admin_alpha — Tue, 20 Oct 2015 23:24:01 +0000

Panagiota Kitsantas, PhD admin_alpha Tue, 10/20/2015 - 19:24

Interim chair, HAP

Contact Information

Email: pkitsant@gmu.edu

Phone: 703-993-1980

CV

Download Panagiota Kitsantas curriculum vitae (CV) here.

Personal Websites

ORCID

Biography

Dr. Kitsantas is a Professor of Biostatistics/Epidemiology in the Department of Health Administration and Policy (HAP). She has also served as the PhD Program Director in Health Services Research at HAP. Previously, she was the Chair of the Department of Population Health at the Schmidt College of Medicine, Florida Atlantic University.

Her research focuses on integrating data science with statistical/epidemiological methods to address health issues in vulnerable populations of women and their children. Dr. Kitsantas is the Principal Investigator of an NIH-funded study examining comorbidities in pregnant Women with prenatal alcohol exposure and adverse birth outcomes. She has also received funding from the Consortium for Medical Marijuana Clinical Outcomes Research to study medical cannabis use among pregnant and non-pregnant women of reproductive age. She has authored and co-authored numerous manuscripts on a wide range of topics, including infant feeding practices, childhood obesity, infant mortality, and various other issues related to maternal health.
Dr. Kitsantas teaches courses in health statistics and research methods, with a strong interest in incorporating AI tools into educational practices to enhance student learning.

Research/Scholarship Interests

Substance use and misuse (e.g., e-cigarettes, alcohol, opioids, and cannabis) among pregnant and non-pregnant women of childbearing age and children, particularly among vulnerable populations such as those with disabilities; machine learning applications in population health; artificial intelligence in teaching and learning.

Select Journal Publications

Matarazzo, A.*, Hennekens, C.H., Dunn, J.*, Mejia, M.C., Levine, R.S., & Kitsantas, P. (2025). New clinical and public health challenges: Increasing trends in United States alcohol related mortality. American Journal of Medicine, 138(3), 477-486.

Kitsantas, P., Benson, K.*, Rubenstein, A.*, Mejia, M. C., Levine, R. S., Hennekens, C. H., & Wood, S. K. (2025). Prenatal cannabis use and adverse health outcomes in neonates and early childhood. Pediatrics and neonatology, S1875-9572(24)00229-8. Advance online publication. https://doi.org/10.1016/j.pedneo.2024.11.004

Kitsantas, P., Densley, S.*, Rao, M.*, Sacca, L., Levine, R.S., Hennekens, C.H., & Mejia, M.C. (2024). Increases in drug-related infant mortality in the United States. Journal of Perinatal Medicine, 52(6), 660-664

Kitsantas, P., Aljoudi, S. M.*, & Sacca, L. (2024). Perception of Risk of Harm from Cannabis Use Among Women of Reproductive Age with Disabilities. Cannabis and cannabinoid research, 9(6), e1615–e1622. https://doi.org/10.1089/can.2023.0199

Yang, J.*, Mejia, M.C., Sacca, L., Hennekens, C.H., Kitsantas, P. (Oct 2024). Trends in marijuana use among adolescents in the United States. Pediatric Reports, 16(4):872-879. doi:10.3390/pediatric16040074

Kitsantas, P., & Pursell, S.R. (2024). Are healthcare providers caring for pregnant and postpartum women ready to confront the perinatal cannabis use challenge? American Journal of Perinatology, 41(S 01):e3249-e3254.

Kitsantas, P., Gimm, G., & Aljoudi, S. M*. (2023). Treatment outcomes among pregnant women with cannabis use disorder. Addictive behaviors, 144, 107723. https://doi.org/10.1016/j.addbeh.2023.107723

Honors and Awards

Recipient of the Shirley S. Travis Habit of Excellence Award, 2014
Recipient of the College of Health and Human Services Master Teacher Award, 2014
Recipient of the �鶹��Ƶ Teacher of Distinction Award, 2014

Professional Affiliations/Memberships

Member, American College of Epidemiology
Member, American Public Health Association

Degrees

PhD, Statistics, Florida State University
MS, Statistics, Florida State University
MS, Health Sciences, James Madison University
BA, Biology, Queens College of the City University of New York