Project: Bioinformatic assessment of multi-center Lipedema transcriptome hallmarks and development of transcriptome-based Lipedema prediction tool with machine learning

Dr. Leon Straub, PhD

Dr. med. Philipp Kruppa

Alan Michael Pittman, PhD

Dhruv Singhal, M.D.

Principal Investigator: Dr. Leon Straub, PhD
Research Faculty, Department of Pharmacology
Department of Pharmacology
University of Virginia
Charlottesville, VA, USA

Co-Principal Investigator: Dr. med. Philipp Kruppa
Attending Physician, Department of Plastic, Reconstructive and Aesthetic Surgery - Hand Surgery and Burn Center
Department of Plastic, Reconstructive and Aesthetic Surgery - Hand Surgery and Burn Center
University Hospital Schleswig-Holstein (UKSH), Campus Luebeck
Luebeck, Germany

Co-Principal Investigator: Alan Michael Pittman, PhD
Department of Molecular & Biomedical Sciences, School of Health and Medical Sciences
City St George's
University of London
London, United Kingdom

Co-Principal Investigator: Dhruv Singhal, M.D.
Department of Surgery
Beth Israel Deaconess Medical Center
Boston, MA, USA

Summary

This research performs the first multi-center transcriptomic analysis of Lipedema, integrating RNA sequencing data from institutions across the United States, Germany, and the United Kingdom. Using advanced bioinformatics and machine learning, this project aims to define Lipedema-specific molecular signatures, distinguish them from obesity and lymphedema, and develop a clinically scalable RNA-based diagnostic prediction tool.

Background

Lipedema is a chronic adipose tissue disorder affecting millions of women, yet it remains frequently misdiagnosed as obesity or lymphedema due to overlapping clinical features. Previous transcriptomic studies have revealed abnormalities in subcutaneous adipose tissue expansion, inflammation, and extracellular matrix remodeling, but were limited by small cohort sizes and single-center designs. These constraints prevented adequate modeling of obesity as a confounding factor or systematic comparison with lymphedema, leaving the molecular boundaries between these conditions poorly defined.

Methodology

This project integrates existing and newly generated RNA sequencing data from subcutaneous adipose tissue biopsies across collaborating international centers.

  • Data Processing: All raw reads are processed through standardized nf-core/rnaseq pipelines aligned to GRCh38. DESeq2 and ComBat-Seq are used for normalization and batch correction, explicitly modeling covariates including BMI, age, menopausal status, tissue location, and sequencing batch.

  • Analysis: Unsupervised clustering methods identify molecular subtypes, while covariate-adjusted differential expression analysis defines Lipedema-specific signatures distinct from obesity and lymphedema.

  • Machine Learning: A three-class machine learning classifier — trained using multinomial elastic-net, XGBoost, random forest, and support vector classification — undergoes nested stratified cross-validation and site-held-out validation to ensure generalizability.

  • Microbial Screening: Parallel exploratory pipelines screen the existing data for non-human RNA signatures to investigate the role of adipose-resident microbial pathogens in disease etiology.

  • Clinical Cohort: Under approved IRB protocols, new tissue samples from 70 patients are collected at BIDMC (led by Dr. Singhal) and at UKSH Luebeck and Bergmann Hospital Potsdam (coordinated by Dr. Kruppa).

Expected outcomes

The investigators hypothesize that Lipedema subcutaneous adipose tissue harbors reproducible molecular subtypes and gene-expression programs distinct from obesity and lymphedema after covariate adjustment. The investigators expect to deliver (1) the first unbiased molecular classification of Lipedema, an obesity-adjusted transcriptomic definition, (2) the first direct transcriptomic comparison between Lipedema and lymphedema, and (3) a validated three-class RNA-based diagnostic classifier with higher accuracy than any existing clinical test.

Practical implementations of results

This research will yield a molecular diagnostic framework that can objectively distinguish Lipedema from obesity and lymphedema, addressing the most common source of clinical misdiagnosis. The resulting machine learning classifier and reduced gene signature will establish a foundation for translating RNA-based diagnostics into clinical practice. Additionally, identification of Lipedema-specific molecular pathways will reveal candidate drug targets for a condition that currently lacks any approved pharmacological treatment.

Our Research Grantees by:   Topic | Recent Projects | Past Projects

LF65_25