A groundbreaking approach to analysing single-cell RNA sequencing (scRNA-seq) data has been introduced by researchers at the National University of Singapore (NUS), led by Associate Professor Zhigang Yao. Named scAMF (Single-cell Analysis via Manifold Fitting), this innovative framework utilises advanced mathematical techniques to fit a low-dimensional manifold within the high-dimensional space where gene expression data are captured. By reducing noise while preserving essential biological information, scAMF enhances the accuracy and speed of data interpretation, holding the potential to accelerate progress in biomedical research, including studies on cancer and Alzheimer’s disease. The collaboration with Professor Yau Shing-Tung from Tsinghua University has resulted in the publication of their findings in the Proceedings of the National Academy of Sciences.
Single-cell RNA sequencing has become a pivotal tool in genomic research, offering unprecedented insights into cellular diversity and disease mechanisms. However, challenges in analysing scRNA-seq data have persisted due to inherent noise from biological variability and technical errors. Traditional analysis methods like genomic imputation, graph-based approaches, and deep learning algorithms often struggle to accurately characterise cell relationships amidst this noise. The scAMF framework represents a significant advancement in overcoming these limitations by denoising scRNA-seq data through manifold fitting, which improves the spatial distribution of the data, bringing gene expression vectors of cells from the same type closer together while maintaining clear separation between different cell types.
Assoc Prof Yao described the innovative approach of scAMF, expressing how the method significantly enhances the accuracy of cell type classification and data visualisation by fitting a low-dimensional manifold in the high-dimensional space. By employing data transformation, manifold fitting using shared nearest neighbour metrics, and unsupervised clustering validation, scAMF outperforms other methods in several key areas such as noise reduction, clustering accuracy, preservation of biological information, computational efficiency, visualisation, and performance across diverse datasets. This positions scAMF as a potent tool in single-cell analysis, potentially revealing previously hidden cellular heterogeneity and rare cell populations.
Building on the success of scAMF, the research team is now focusing on developing a novel framework for constructing high-resolution, multiscale cell atlases to overcome current methodological limitations in cell atlas construction. This new approach aims to address challenges in identifying small cell populations and outdated unsupervised learning techniques. By developing a multi-resolution cell analysis framework based on scAMF, researchers aim to identify rare cell populations and contribute to the construction of comprehensive cell atlases, allowing for the analysis of cellular heterogeneity at various levels of granularity, from broad cell types to subtle subpopulations. This is especially crucial for identifying rare cell types that may be overlooked by conventional analysis methods.
Assoc Prof Yao highlighted how ongoing research has demonstrated promising results across benchmark datasets, unveiling novel biological insights when applied to the Human Brain Cell Atlas to identify new subtypes and marker genes for various cell types. This ongoing work holds the promise of pushing the boundaries of single-cell analysis further, potentially revolutionising understanding of cellular diversity and function across various biological systems. By employing scAMF and developing a multi-resolution cell analysis framework, researchers aim to drive greater understanding of cellular diversity and function in biomedical research, ultimately leading to advancements in the field of genomics and precision medicine.
