報告題目:scFMA: A Flexible Random Effects Model for DNA Methylation Analysis with scBS-seq Data
主講人:黃希芬教授(云南師范大學)
時間:2025年10月27日(周一)15:00 p.m.
地點:北院卓遠樓305會議室
主辦單位:統計與數學學院
摘要:Background: Single-Cell Bisulfite Sequencing (scBS) has advanced rapidly because of its ability to study DNA methylation patterns at the level of individual cells. Since DNA methylation influences gene expression, a common approach involves identifying regions of differential methylation between cells to further explore transcriptional regulatory features and intercellular heterogeneity. However, this technique faces challenges such as limited sequencing depth and coverage, difficulties in processing large numbers of missing values, and the prevalence of extreme methylation rates (0 and 1) in single-cell data. These factors complicate accurate differential methylation analysis between cells.
Results: We develop a binomial model with flexible random effects for DNA methylation analysis. By combining the discrete approximation technique with the MM algorithm for model estimation and using the bootstrap method to construct a test statistic for model testing, the proposed scFMA method identifies differentially methylated regions without restricting the form of the random-effects distribution, which solves the limitation that the existing methods are prone to model misspecification that leads to estimation bias.
Conclusion: Numerical studies have demonstrated that scFMA can identify more differentially methylated regions while maintaining low false positive rates across various distribution settings. We applied this method to investigate the mechanisms of cell methylation in early embryos. Despite the challenges posed by sparse data and small sample sizes, our model effectively captures the actual data distribution and reveals critical biological information during embryonic development.
主講人簡介:
黃希芬,云南師范大學數學學院教授,主要從事生物醫學中統計模型和快速算法等方面的研究,主持國家自然科學基金項目3項,發表論文20余篇。2019年和2023年先后獲得云南省“興滇英才支持計劃”青年人才和云南省社會科學獎等獎項。