ML models infer tumor features from cfDNA fragmentomics and single-cell data

Published papers

• scMILD: Single-cell multiple instance learning for sample classification and associated subpopulation discovery.

  iScience2026-03-10Kyeonghun Jeong et al.

  scMILD is a weakly supervised multiple instance learning framework that identifies condition-associated cells from sample-level labels in single-cell data,…

• Fragmentomic liquid biopsy enables early breast cancer detection, molecular subtyping and lymph node assessment.

  Nature communications2026-03-06Yuxuan Zhu et al.

  TuFEst, an ML model on genome-wide cfDNA fragmentomics, detects early breast cancer, subtypes tumors and predicts lymph node status.

Preprints

• PROTOTYPE-BASED CONTINUAL LEARNING FOR SINGLE-CELL ANNOTATION

  bioRxiv2026-03-08Ge, S. et al.

  scEvolver is a continual learning framework for single-cell annotation that accumulates cell-type prototypes without revisiting historical data, enabling…

• Classification with Missing Data - A NIFty Pipeline for Single-Cell Proteomics

  bioRxiv2026-03-09Nitz, A. A. et al.

  NIFty is a top-scoring pairs feature selection and classification pipeline for single-cell proteomics that handles missing data, avoids double dipping, and…

• Exploring per-base quality scores as a surrogate marker of cell-free DNA fragmentome

  bioRxiv2026-03-10Volkov, H. H. V. et al.

  Per-base quality scores in cfDNA WGS carry fragmentomic signal; a leave-one-batch-out classifier reaches pooled AUC 0.81 for cancer.

Related topics

• Quantitative Models In Cell Biology
• Cross Modal Spatial Omics Inference
• Personalized Biomedical Modeling Frameworks

Topic history