Research
We study how aging shapes somatic genomic variation and multi-omic profiles—and how these changes contribute to cardiovascular and other age-related diseases. We develop and apply computational methods that integrate human multi-omics with medical imaging, EHR data, and experimental models to identify causal disease mechanisms, biomarkers, and genetically supported therapeutic targets.
Across our research, we aim to connect molecular variation to human phenotypes and disease, and ultimately to opportunities for prevention and treatment.
Aging, somatic mosaicism, and disease
Somatic mutations accumulate throughout life, creating genetically distinct cell populations in blood and other tissues. We study the factors that shape these mutations and their consequences for human health, with a particular focus on clonal hematopoiesis.
By integrating population-scale genetics, transcriptomics, proteomics, clinical phenotypes, and human and murine studies, we investigate how somatic mutations contribute to cardiovascular, kidney, inflammatory, infectious, and malignant diseases. Beyond identifying associations, we triangulate evidence across these modalities to define causal mechanisms and prioritize therapeutic targets that may disrupt the progression from aging to somatic mutation to disease.
Selected publications and preprints:
- Somatic mosaicism across tissues: SMaHT Network, Cell, accepted in principle; Ha et al., Cell Genom, accepted; SMaHT Network, bioRxiv, 2026; Coorens et al., Nature, 2025; and Yu, Coorens et al., Nat Rev Genet, 2024.
- Genomic and molecular determinants: Uddin, Yu et al., medRxiv, 2026; Yu et al., Nat Commun, 2025; and Yu et al., J Clin Invest, 2023.
- Cardiovascular and metabolic disease: Liu et al., JACC CardioOncol, accepted in principle; Yalcinkaya et al., Circulation, 2026; Chang et al., Eur Heart J, 2026; and Zuriaga, Yu et al., Eur Heart J, 2024.
- Kidney, inflammatory, infectious, and malignant disease: Fewings et al., Nat Commun, accepted; Kapur, Li et al., J Clin Invest, 2026; Kumar, Li et al., Blood, 2026; and Yu et al., J Am Soc Nephrol, 2025.
Multimodal phenotyping and machine learning
Important aspects of human biology are captured not only by molecular measurements, but also by medical images and longitudinal clinical records. We develop machine-learning approaches to derive quantitative phenotypes from medical imaging and use large language models to interpret complex information from EHRs.
We then integrate these phenotypes with genomic and other omics data to uncover biological mechanisms and improve disease characterization. Current applications include splenic, cardiac, and liver imaging; biological-age estimation; and multimodal prediction of cardiovascular and other age-related diseases.
Selected publications and preprints:
- Medical imaging: Kamineni et al., Sci Transl Med, 2026; Tian et al., medRxiv, 2026; and Zekavat et al., Circulation, 2022.
- Clinical risk prediction: Yu et al., medRxiv, 2022.
Omics and statistical methods
Large-scale omics studies introduce fundamental methodological challenges, including measurement error, discordance across experimental platforms, detection of low-frequency somatic mutations, and uncertainty in the clinical interpretation of genetic risk.
We develop computational and statistical methods to address these challenges. Current areas include benchmarking somatic-mutation detection, cross-platform proteomic imputation and quality assessment, evaluating the individual-level performance of polygenic risk scores, and integrating genetic instruments with molecular biomarkers for causal inference and therapeutic target prioritization.
Selected publications and preprints:
- Proteomics: Li, Alaa et al., bioRxiv, 2026; and Katz et al., Sci Adv, 2022.
- Polygenic risk: Liang, Kim et al., Nat Commun, accepted in principle; Zhang et al., Nat Genet, 2023; and Yu et al., J Am Soc Nephrol, 2021.
- Causal inference: Jin et al., Biostatistics, 2024; and Yu et al., Kidney Int, 2020.
How we work
Our work is grounded in large and diverse human cohorts, biobanks, clinical datasets, and collaborative experimental studies. We combine population-scale discovery with mechanistic investigation in human and murine systems, working closely with clinicians, experimental scientists, cohort investigators, and consortium partners.
