The Molecular Stratification of EDS (MoSEDS) Framework

MoSEDS operates as a three-layer pipeline. Each project performs a distinct analytical function, and together they form a progression from genetic variant interpretation to pain mechanism annotation to integrated patient subtype classification.

The three layers follow an order: KLKVar and ACKRScope are parallel first-order tools that both feed into HyphenScore. KLKVar characterizes the ECM-remodeling genetic burden from the KLK protease family. ACKRScope characterizes the pain-amplification genetic burden from the ACKR3 locus. HyphenScore integrates both alongside clinical phenotype data to produce biologically annotated patient subtype classifications.

KLKVAR (Layer 1A)

• KLKVar is a composite pathogenicity scoring pipeline for missense variants across all 15 KLK genes. The KLK gene family encodes a cluster of serine proteases whose job is controlled remodeling of extracellular matrix proteins (think of them as molecular scissors for connective tissue maintenance). When variants in these genes produce abnormal protease behavior, the connective tissue demolition process runs unchecked relative to construction, and tissue architecture degrades. KLKVar applies the same analytical architecture as ChanVar (the CACNA1A variant pipeline from MiSOF Layer 1): nine quantitative features combined into a Composite Pathogenicity Score with bootstrap confidence intervals. Features include allele frequency, thermodynamic stability, evolutionary conservation, active site proximity, substrate interaction domain, and tissue expression context. The tool is validated against published WES data from 197 hEDS patients.

ACKRScope (Layer 1B)

• ACKRScope annotates variants in the ACKR3 locus identified by the 2025 hEDS genome-wide association study. ACKR3 is an atypical receptor with an unusual function: it sequesters endogenous opioid peptides (i.e. enkephalins, dynorphins, the body's own pain-relief molecules) and removes them before they can reach the receptors that would reduce pain. It is, effectively, a drain that catches your own painkillers.

• The 2025 GWAS found that hEDS risk variants near ACKR3 predict increased ACKR3 expression in peripheral nerve tissue. More ACKR3 expression means more opioid scavenging, which means a genetically amplified chronic pain burden on top of the structural joint problems. ACKRScope takes a patient's genotype in the ACKR3 region and returns an ACKR3 Expression Risk Score: a predicted pain amplification index grounded in eQTL data, enhancer activity annotation, and transcription factor motif analysis. It's the first tool to translate this GWAS finding into a patient-facing score.

HyphenScore (Layer 2)

• HyphenScore is the integrative layer. It takes genetic scores from KLKVar and ACKRScope, combines them with clinical phenotype data, and applies hierarchical clustering to identify biologically annotated patient subtypes within the hEDS population.

• hEDS almost certainly isn't one disease. It's a phenotypically similar cluster of conditions arising from different molecular mechanisms. The Norris lab at MUSC identified three distinct phenotypic clusters in 2,149 hEDS patients using k-means clustering, but didn't establish what biologically drives each cluster. HyphenScore extends that work: it attaches genetic pathway interpretation to each cluster using GWAS-derived genetic correlation data from 19 comorbid conditions, producing provisional labels like "ECM-remodeling dominant" or "neuroimmune amplification dominant." This matters for drug development. For instance, a KLK inhibitor will only work in patients whose hEDS is KLK-driven. An ACKR3 modulator will only work in patients with elevated pain amplification. HyphenScore is the patient stratification infrastructure that makes targeted clinical trials possible.

You can think of MoSEDS as a system for figuring out which version of hEDS a given patient has, and why their particular combination of symptoms makes biological sense.

The first step (KLKVar) looks at the genes (recently discovered to be KLK genes) responsible for maintaining connective tissue. If someone has rare variants in those genes, the tool scores how likely those variants are to be causing the tissue problems associated with hEDS. The second step (ACKRScope) looks at a different system: the body's pain regulation machinery. Some people have genetic variants that predict their nervous system scavenges natural pain relief more aggressively than average. This doesn't cause the structural joint problems, but it does explain why the pain burden is often so much worse than the structural findings would suggest in hEDS individuals. 

The final step (HyphenScore) takes both of those scores together, along with a patient's full symptom picture, and asks: which biological subtype does this person most likely represent? This answer matters because different subtypes probably respond to different treatments, and building that stratification layer now is what makes better targeted therapies or drugs possible later.

And of course, none of this treats patients directly. MoSEDS analyzes data and builds interpretive tools with the hope that these tools help researchers ask better questions and eventually help clinicians make better decisions.