dsLassoCov - A federated machine learning approach incorporating covariate control

Aug 2, 2026

Upload by Ophelia

Machine learning in biomedical research increasingly relies on large datasets, yet cross-institution integration is hindered by legal restrictions and complex data governance. Federated learning enables privacy-preserving model training across geographically distributed data, but remains challenged by controlling covariate effects under statistical heterogeneity. Conventional covariate-control strategies incur prohibitive communication costs, especially for high-dimensional settings. dsLassoCov targets efficient covariate-effect control in federated learning, supporting biomarker selection while reducing confounding. Simulations and multi-database exposome analyses validate robust performance consistent with prior studies.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages35

Reading Time1 h 28 min

AccessFree to download

Abstract

Introduction

Federated learning in biomedical cross-silo scenarios

Statistical heterogeneity and covariate effects

dsLassoCov: a federated machine learning approach incorporating covariate control Han Cao\#a, Augusto Anguita\#b,h, Charline Warembourgc, Xavier Escribà-Montagutb,h , Martine Vrijheidb,h, Juan R. Gonzalezb,h, Tim Cadmanb,g, Verena Schneider-Lindnerd, Daniel Durstewitza, Xavier Basagañab,h*, Emanuel Schwarze,f* \# These authors contributed equally to this work * Corresponding authors a Department of Theoretical Neuroscience, Central Institute of Mental Health, Medical Faculty, Heidelberg University, Germany b ISGlobal, Barcelona, Spain. c Univ Rennes, Inserm, EHESP, Irset (Institut de recherche en santé, environnement et travail) -UMR_S 1085, F-35000 Rennes, France. d Department of Anesthesiology and Surgical Intensive Care Medicine, Medical Faculty Mannheim, Heidelberg University, TheodorKutzer-Ufer 1-3, 68167, Mannheim, Germany e Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany f Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty, Heidelberg University, Mannheim, Germany g Genomics Coordination Center, Dept of Genetics, University Medical Centre Groningen, Groningen, Netherland h Universitat Pompeu Fabra (UPF), Barcelona, Spain; CIBER Epidemiologa y Salud Pública, Madrid, Spain. Abstract Machine learning has been widely adopted in biomedical research, fueled by the increasing availability of data. However, integrating datasets across institutions is challenging due to legal restrictions and data governance complexities. Federated learning allows the direct, privacy preserving training of machine learning models using geographically distributed datasets, but faces the challenge of how to appropriately control for covariate effects. The naive implementation of conventional covariate control methods in federated learning scenarios is often impractical due to the substantial communication costs, particularly with high-dimensional data. To address this issue, we introduce dsLassoCov, a machine learning approach designed to control for covariate effects and allow an efficient training in federated learning. In biomedical analysis, this allow the biomarker selection against the confounding effects. Using simulated data, we demonstrate that dsLassoCov can efficiently and effectively manage confounding effects during model training. In our real-world data analysis, we replicated a large-scale Exposome analysis using data from six geographically distinct databases, achieving results consistent with previous studies. By resolving the challenge of covariate control, our proposed approach can accelerate the application of federated learning in large-scale biomedical studies. Introduction Big data technologies have been widely adopted in biomedical research, with machine learning (ML) analysis demonstrating significant success in various molecular1 and clinical2 applications. This success stems from the development of accurate and generalizable models through training on extensive datasets. However, a central challenge in biomedical research is that data sources are often confined to isolated databases within individual institutes or hospitals. These databases often operate under distinct standards and governance structures3, limiting the ability to pool data effectively. The impediment to data pooling is also made difficult by policy regulations, such as the General Data Protection Regulation4 (GDPR) in the EU, which requires explicit consent from the patients for the transfer of their data3. This challenge significantly hampers the generalizability of ML technologies, e.g., deep learning, which heavily relies on access to extensive training data, and becomes more pronounced in data-scarce areas, such as rare disease analysis5. As highlighted in a recent review6, federated learning emerges as a promising cross-silo ML technology that safely facilitates mode

FAQs

What problem does dsLassoCov address in federated learning?

dsLassoCov addresses how to control covariate effects during federated model training despite statistical heterogeneity across datasets.

Why are conventional covariate-control methods difficult to use in federated learning?

They often require substantial communication, making naive implementations impractical, particularly with high-dimensional data.

How is dsLassoCov evaluated in the document?

It is evaluated using simulated data to test confounding control, and using real-world exposome analysis across six geographically distinct databases to confirm results consistent with previous studies.

Aug 2, 2026

Upload by Ophelia