(Boston)—Hypermobile Ehlers–Danlos syndrome (hEDS) is one of the most common heritable connective tissue disorders. Early estimates have reported that this genetic disorder affects at least one in 5,000 individuals and more recently it has been estimated to affect upwards of 1-3% of the population worldwide. Clinically, hEDS is characterized by generalized joint hypermobility, tissue fragility including capillary fragility associated with easy bruising, poor wound healing and atrophic scarring, and skin hyper-extensibility. A particularly concerning complication of hEDS that has been underrecognized is the occurrence of fragility fractures in infancy and childhood and the social and legal consequences that can result from diagnostic errors.
Despite the clinical recognition of hEDS for decades and advances in genetic sequencing technologies, the molecular basis of hEDS has remained largely mysterious until now. Using for the first time machine learning with rigorous subject-level statistical analysis, researchers from Boston University Chobanian & Avedisian School of Medicine believe that hEDS is not a single-gene disorder, but rather involves a combination of genetic variations affecting three key biological systems. They stress that these biological associations represent baseline genetic data that prioritize hypotheses for future in-depth investigation, rather than established disease mechanisms.
"Hypermobile Ehlers–Danlos Syndrome represents 80-90% of all EDS cases, yet the vast majority remain undiagnosed due to lack of awareness among healthcare providers and the absence of a definitive genetic test," explains corresponding author Michael F. Holick, PhD, MD, professor of medicine, pharmacology, physiology & biophysics and molecular medicine at the school. "Through systematic clinical phenotyping and genomic analysis of affected families, we have aimed to uncover the genetic architecture underlying hEDS and translate these findings into improved diagnostic and therapeutic strategies."
Using Whole Exome Sequencing to identify potential functional mutations, the researchers analyzed 35,923 rare genetic variants of 116 people from 43 families. This included 86 patients diagnosed with hEDS and 30 unaffected family members as controls. They then used machine learning including random forest (for picking the most important genes), deep neural networks and ensemble methods (combining multiple models) to identify patterns in the data.
Their analysis found that individuals with hEDS have a significantly higher burden of variants (genetic differences) in three distinct biological areas: the collagen biosynthesis pathway variants; the HLA/adaptive immune axis variants and the mitochondrial respiratory chain variants compared to a control group, providing a potential polygenic (multi-gene) explanation for the disorder.
According to the researchers, these findings have potential implications for improved genetic counseling, risk stratification of affected families, and future development of precision medicine approaches tailored to individual genetic profiles. "More broadly, this work demonstrates a successful application of integrated machine learning to genetic variant discovery in a complex, genetically heterogeneous rare disease, providing a methodological template that may be applicable to other missing heritability problems in human genetics."
These finding appear online in the journal Genes.
