Related Experiment Video
Updated: Jun 8, 2025
03:37
Author Spotlight: Impact of Intergenic Interactions on Disease-Identifying Dark Biomarkers
Published on: March 1, 2024
645
HiDDEN: a machine learning method for detection of disease-relevant populations in case-control single-cell
Aleksandrina Goeva1, Michael-John Dolan2, Judy Luu2
1Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge, MA, USA. aleksandrina.goeva@utoronto.ca.
Nature Communications
|November 2, 2024
Summary
Standard single-cell RNA-seq analysis misidentifies affected cells. A new method, HiDDEN (Hidden), accurately refines cell labels to detect subtle biological signals in case-control studies.
Area of Science:
- Computational Biology
- Genomics
- Immunology
Background:
- Case-control single-cell RNA-seq studies often incorrectly label all cells in a case sample as perturbed.
- This standard approach fails to identify subtle or small subsets of affected cells and their specific markers.
Purpose of the Study:
- To introduce HiDDEN, a novel computational method for refining cell-specific labels in case-control single-cell RNA-seq data.
- To demonstrate HiDDEN's capability in accurately identifying perturbed cells and biological signals missed by conventional methods.
Main Methods:
- Simulations were used to evaluate the performance of the standard analysis versus HiDDEN.
- HiDDEN was applied to datasets from human multiple myeloma precursor conditions and a mouse model of demyelination.
Main Results:
- HiDDEN successfully recovered subtle biological signals in simulated datasets.
- In human multiple myeloma, HiDDEN identified early-stage malignancy missed by original analysis.
- In a mouse demyelination model, HiDDEN identified an endothelial subpopulation involved in blood-brain barrier dysfunction.
Conclusions:
- HiDDEN offers a superior approach to standard single-cell analysis for detecting subtle transcriptional changes.
- The method accurately refines cell labels, improving the identification of affected cells and biological insights across various research contexts.
