Talk Title:

Modeling and Inference with Noisy Biological Data to Study Cellular Mechanisms

Abstract: 

Biological data, including data produced in wet labs and data obtained with computational predictions, tend to be highly noisy. It is crucial to model and reduce the noise before using the data to answer specific biological questions. Single-cell RNA-sequencing data provides whole-transcriptome information with unprecedented scale and resolution, but this data is confounded by multiple noise factors. We developed a simulator SymSim to model the biological and technical processes that generate single cell RNA-seq data, providing parameters to control the amount of each noise factor introduced during these processes. We then used this simulator to benchmark computational methods for downstream analysis, and investigated the effect of each noise factor on the performance of the methods. We can also use this simulator to assist wet-lab experimental design.
 
One application of single-cell RNA-seq data is the inference of transcriptional regulatory networks, which are high-level representation of regulatory mechanisms in cells. However, there is usually a considerable amount of noise (false positive and false negative interactions) in the resulting networks. We developed an evolutionary approach with a probabilistic graphical model to reduce the noise in the networks of a collection of species. This evolutionary framework can be applied to other mechanisms learned with data on single cell level.

Biography: 

Xiuwei Zhang is a postdoctoral researcher in Prof. Nir Yosef's group at UC Berkeley. She obtained a PhD in computer science from EPFL (École Polytechnique Fédérale de Lausanne), Switzerland. Before moving to the United States, Xiuwei was a postdoc researcher in Dr. Sarah Teichmann's group, at the European Bioinformatics Institute (EBI) and Wellcome Trust Sanger Institute in Cambridge, UK. She was a research fellow of the 2016 program on Algorithmic Challenges in Genomics at the Simons Institute for the Theory of Computing.