报告题目：Functional Genomic Data Analysis and Integration

报告人：Guan Yuanfang,Assistant Professor, Department of Computational Medicine & Bioinformatics, University of Michigan

报告时间：2013年11月12日 上午10：00-12：00

报告地点：中国科学院遗传与发育生物学研究所B210会议室

摘要：

In this talk, we present a novel network time-course projection method that achieved the best performance (along another team) in DREAM8 subchallenge 2A, as well as the best aggregate prediction to subchanllenges 2A and 2B: Time course prediction in breast cancer cell lines. Additionally, we will present a method that integrates RNA-seq data to predict functions at the alternatively spliced isoform level. Perturbation experiments are vitally important in causal signaling network inference, drug treatment simulation, etc. However, such experiments are expensive and time-consuming. How to generalize the observed time-course network dynamics to unseen situations remains a challenging task. To solve this problem, we developed a protein phosphorylation dynamics prediction method using truncated singular value decomposition (SVD). Any time-course data could be used as inputs (with or without inhibitors) to predict the perturbation under other inhibitors within the same cell culture. In addition, we will present a multiple-instance-learning based method to predict alternatively spliced isoform functions. We show that our method can successfully assign functions at the isoform level and also improves gene-level prediction performance.