Analysis and modeling of single-cell genomics data to drive biomedical discovery and innovation
发布时间 :2019-10-17  阅读次数 :4095

报告人:Dr. Jun Ding,Carnegie Mellon University

时间:2019年10月19日(星期六),10:30-11:30

地点:闵行校区生物药学楼3-117会议室

联系人:魏冬青、熊毅

 

报告人简介:

Jun Ding (http://www.cs.cmu.edu/~jund/) is a Postdoctoral fellow in the Computational Biology Department, School of Computer Science, Carnegie Mellon University under the supervision of Dr. Ziv Bar-Joseph. Previously, he received his PhD. in Computer Science from the University of Central Florida. His research focuses on developing computational methods to drive biological discoveries and medical innovations that will benefit public health by analyzing and modeling large-scale biomedical data, especially single-cell genomics data. Jun had published over 20 journal papers in leading computational biology journals such as genome research and cell stem cell. Many computational models he developed have significantly contributed to the studies in the areas that he worked on. For example, Jun has improved the cell differentiation protocol from iPSCs to lung epithelial cells with the single-cell differentiation models he developed.  Also, Jun has developed one of the very first machine learning based microRNA target prediction methods that learn the new canonical microRNA binding features from high-throughput microRNA binding experiments to improve the prediction accuracy. Currently, Jun is particularly interested in developing computational models and visualizations for cell differentiation and tumor microenvironment studies that enormously benefit from single-cell technologies.

 

报告摘要:

In recent years, the emerging single-cell technologies provide unprecedented opportunities for studying many challenging biomedical problems, especially in the cell differentiation and cancer biology areas, in which there exists tremendous cell heterogeneity.  However, the single-cell datasets generated in those studies are usually high-dimensional, large-scale, and noisy, which makes it challenging for the biomedical scientists to directly utilize the single-cell data in support of their health science researches. In this talk, I will discuss how to make novel biological discoveries or medical innovations in the cell differentiation studies that will benefit public health by computationally analyzing, modeling, and visualizing large-scale single-cell genomics datasets.  First, I will present a novel computational method (Kalman filter based) on reconstructing cell differentiation trajectories and the underlying regulatory networks from the time-series single-cell RNA-seq data.  Unlike other existing single-cell expression based trajectory inference methods, the proposed model can deal with the dropout noise very efficiently, and it is one of the very first methods that can infer a list of cell fate dictating regulators for the differentiation process. Besides, the model has incorporated interactive visualization functions, which can dynamically generate appropriate visualizations based on the interests of the users. With the computational model, we have discovered and experimentally verified many novel regulators for the lung and heart developmental processes. Second, I will talk about how to fully exploit the single-cell RNA-seq data to reconstruct the cell differentiation trajectories better. Most existing methods are only using the gene expression information from the single-cell RNA-seq data, and other information associated with those single-cell RNA-seq is neglected. Here, I have shown that such neglected information (e.g., mutations) can be very informative for trajectory inference. The cell differentiation trajectories can be significantly improved by combing the “mutations” and gene expression information from the single-cell RNA-seq data together.  Lastly, I will talk about how to innovate the start-of-the-art protocol for differentiating iPSCs to lung epithelial cells by computationally modeling the single-cell RNA-seq data. With a careful designing of the single-cell experiment and the following computational modeling and visualization of the generated single-cell genomics datasets, we have discovered a significant reason for the cell fate divergence in the studied differentiation process. By repressing the identified pathway at the predicted time point from our model, we have dramatically improved the efficiency of the start-of-the-art differentiation protocol that will significantly benefit the lung regenerative medicine studies.