Related Experiment Video
Updated: Aug 12, 2026
A Versatile Automated Platform for Micro-scale Cell Stimulation Experiments
Published on: August 6, 2013
SKIM: A fast sketching strategy integrated with model's dynamic-feedback for large-scale single-cell transcriptomic
Jiaxing Bai1, Feng Zhou2, Chongyang Tan2
1Department of Automation, National Institute for Data Science in Health and Medicine, State Key Laboratory of Mariculture Breeding, Xiamen Key Laboratory of Big Data Intelligent Analysis and Decision, Xiamen University, Xiamen, Fujian, China.
SKIM is a novel method for single-cell RNA sequencing (scRNA-seq) data analysis that efficiently sketches large datasets. It uses dynamic feedback to identify abnormal cells and balance populations, improving downstream analysis and reducing computational burden.
Area of Science:
- Genomics
- Computational Biology
- Bioinformatics
Background:
- Single-cell RNA sequencing (scRNA-seq) generates massive datasets, leading to computational challenges and cell population imbalances.
- Existing dataset sketching methods use geometric distances, failing to capture local expression variations and overselecting peripheral cells.
- There is a need for efficient sketching methods that preserve biological signals while handling large-scale scRNA-seq data.
Purpose of the Study:
- To introduce SKIM, a fast sketching strategy integrated with dynamic feedback for large-scale scRNA-seq analysis.
- To overcome limitations of existing methods by capturing multiscale transcriptomic differences and reducing computational burden.
- To improve cell type annotation, data integration, deconvolution, and trajectory inference.
Main Methods:
- SKIM utilizes dynamic feedback, defined as reconstruction losses across training epochs, to capture multiscale transcriptomic variations.
- Abnormal cells with unstable feedback patterns are identified and removed.
- Sketches are constructed via clustering and size-aware sampling in the dynamic-feedback space, balancing cell populations.
Main Results:
- SKIM outperforms four state-of-the-art sketching methods across nine benchmark datasets.
- The method shows significant improvements in cell type annotation, data integration, bulk RNA-seq deconvolution, and developmental trajectory inference.
- SKIM achieves over a 20-fold speedup compared to existing methods for large-scale dataset sketching.
Conclusions:
- SKIM provides an efficient framework for large-scale scRNA-seq dataset sketching.
- The dynamic feedback approach effectively retains critical biological signals while reducing computational burden and balancing cell populations.
- SKIM enhances the analytical capabilities for large scRNA-seq datasets, facilitating deeper biological insights.
