University of South Florida

Researchers are using data mining to learn more about diabetes cases that don’t fit the usual labels

In the ongoing research and treatment of diabetes, the focus is typically on the two forms of the disease that dominate public awareness. Type 1 is caused by the immune system attacking the cells that produce insulin, so that the person requires insulin therapy for life; type 2 is frequently associated with obesity and lack of exercise, resulting in insulin resistance.

But researchers are learning more about patients whose symptoms are classified as “atypical,” thanks to the Rare and Atypical Diabetes Network (RADIANT), led by teams at USF Health, Baylor College of Medicine, the University of Chicago, and Massachusetts General Hospital. The USF Health research team is led by Jeffrey Krischer, Ph.D., director of the USF Diabetes and Endocrinology Center and the USF Health Informatics Institute, and includes Hemang Parikh, Ph.D., an associate professor in bioinformatics and biostatistics in the Health Informatics Institute.

Their paper, “Data Mining Framework for Discovering and Clustering Phenotypes of Atypical Diabetes,” was recently published in the Journal of Clinical Endocrinology & Metabolism in collaboration with Ashok Balasubramanyam, M.D., and Maria Redondo, M.D., Ph.D., from Baylor College of Medicine, and Christiane Hampe, Ph.D., from the University of Washington.

Hemang Parikh, PhD, of the USF Health Informatics Institute

“In addition to type 1 and type 2 diabetes, there is a range of atypical forms of diabetes that affect people who cannot be categorized in the same way,” said Dr. Parikh. “Sometimes, these people – children and adults – are misdiagnosed and receive different treatment than they should get.”

One form of atypical diabetes – monogenic– is due to the dysfunction of a single gene. Another type results from a cluster of genetic disorders and relates to abnormal mitochondrial function. Another is characterized by patients who appear to have type 2 diabetes, yet present with diabetic ketoacidosis, a complication thought to occur only in patients with type 1 diabetes. Yet another affects the manner in which fat is stored.

The new paper furthers the study of these rarer forms of the disease, for which patients’ symptoms and health challenges differ from those with type 1 and type 2. The analysis was conducted through the sophisticated process of data mining – digging through data to discover hidden patterns and is sometimes referred to as “knowledge discovery in databases.”

Dr. Parikh and his team developed a data mining system as part of a program called DiscoverAD (short for Discover Atypical Diabetes). In essence, DiscoverAD relies on a two-step filtering process – first to exclude participants who meet definitions of the typical type 1 diabetes or type 2 diabetes, then to include participants with certain pre-specified atypical diabetes characteristics.

“This is followed by robust analysis to discover novel phenotypes of atypical diabetes (AD) within the filtered group,” Cassandra Remedios, M.S., an assistant in research in bioinformatics in the Health Informatics Institute said. “We purposefully developed DiscoverAD to permit flexibility and efficiency so it can be applicable to various clinical settings with different types of large cohort datasets.”

In the study, two distinct and very different cohorts of patients with diabetes were investigated. The first cohort comprised Hispanic participants with diabetes from the Cameron County Hispanic Cohort led by Joseph McCormick, M.D., and Susan Fisher-Hoch, M.B.B.S., M.Sc., F.R.C. Path., M.D., of the University of Texas Health Sciences Center. The second cohort comprised 758 multiethnic children within the Texas Children’s Hospital Registry for New-Onset Type 1 Diabetes study. The investigation allowed them to identify and cluster phenotypes of atypical diabetes. “Due to the large cohort datasets, a manual review would have been extremely time-consuming,” Parikh explained.

The study was conducted as part of RADIANT, which is dedicated to discovering and defining rare and atypical forms of diabetes. RADIANT is comprised of universities, hospitals, and clinics around the United States working to gain a deeper understanding of atypical diabetes. Baylor College of Medicine and the University of Chicago are the national centers of the consortium, and USF serves as the data coordinating center for the entire network.

“This work demonstrates the high prevalence of atypical forms of diabetes in varied populations. The DiscoverAD tool is an innovative and practical tool to identify such patients in different datasets. I believe this could be a foundation for developing criteria that clinicians can use to diagnose their patients with diabetes more accurately and treat them more precisely,” Dr. Balasubramanyam said.

The idea of RADIANT’s study, which is funded by the National Institute of Diabetes and Digestive and Kidney Diseases, part of the National Institutes of Health, is to identify people who have atypical diabetes – which is a cluster of traits, “ Dr. Parikh explained. “It can’t be characterized in just one way. There is heterogeneity in atypical cases.”

The significance of the research lies in the fact that some patients with rare forms of diabetes remain undiagnosed or could possibly receive incorrect treatment. Proper diagnosis, according to RADIANT, “enables targeted therapy, leads to improved quality of life, and aids in the diagnosis of diabetes in other family members.”

Patients with atypical diabetes are treated throughout the country, but frequently as isolated, individual cases, and that has made it difficult to amass a base of knowledge that benefits providers and patients. RADIANT addresses that challenge by creating a centralized database, information, and resources – with the goal of leading to more effective diagnoses and better treatment plans.

“We found in our studies that atypical cases are quite high – comprising about five to eleven percent of diabetes diagnoses,” Dr. Parikh said. “And we also found that many people have been misdiagnosed as either type 1 or type 2 diabetes.”

A key indicator of atypical diabetes is a treatment that does not seem to be working. For instance, some diabetes patients might start losing weight quickly and inexplicably. Others may see glucose levels remain high despite receiving insulin.

“If a person is not responding in a way they should be, that could be a sign,” Dr. Parikh said.

As one of the nation’s leading diabetes researchers, Dr. Krischer has been studying data related to atypical diabetes for several years. Several hundred subjects have been involved in the RADIANT study to gain a greater understanding of atypical diabetes through data mining.

Jeffery Krischer, PhD, at the USF Health Informatics Institute on USF’s campus in Tampa.

 “These new analytic techniques make it easier to recognize atypical forms of diabetes that can lead to better management tailored to individual case’s characteristics,” said Dr. Krischer, who also holds the USF Health Endowed Chair in Diabetes Research. “Not only does this demonstrate the potential of personalized medicine, but the analytics also define computable phenotypes that can be generalized to many data mining situations.”

Dr. Parikh, whose expertise lies in biomedical data analysis, has a personal connection to the study of diabetes. Both of his late grandfathers developed type 2 diabetes and his mother has it as well. That led him to want to learn more about the disease.

He finds his work with RADIANT particularly rewarding, with a collaborative approach to researching and treating atypical diabetes.

“It’s a huge consortium, and we’re the main data coordinating center, which means we focus on the aggregation and management of data from multiple sites,” he said. “Then we work with these different network partners for processing of the biospecimens for high-throughput technologies such as whole-genome sequencing, RNA-sequencing, or metabolomics and deep phenotyping of individuals and/or families.”

The study is ongoing, Dr. Parikh stresses. If somebody suspects they may have atypical diabetes or know someone who might, they can visit the RADIANT website ( Visitors are asked to complete a questionnaire and then, based on the responses, could be enrolled in the study.

“There is a considerable number of atypical cases out there that people may not realize,” Dr. Parikh said. “And we want those people to receive the proper diagnosis, to avoid causing diabetes complications over time.”

— Story by Dave Scheiber for USF Health Communications 

Network-wide options by YD - Freelance Wordpress Developer