AISEM series provide a unique opportunity for both new and returning students to interact with groups of the UW Artificial Intelligence community (including machine learning, bioinformatics, computer vision etc.) and hear about their research.

Each seminar will feature a unique group. The seminar includes several short talks given by students in that group and a social time which the PI will be present. Refreshments are provided.

In this seminar, we will introduce Prof. Sushmita Roy’s lab. The Roy lab is interested in developing and applying machine learning algorithms to understand the structure, function and evolution of regulatory networks. Her lab develops methods to integrate diverse regulatory genomics datasets to identify such networks. Her lab is also interested in understanding how these networks change across different biological contexts (e.g. different time points, different species), and how this affects the behavior of organisms.

Prof. Roy and one of her student Rob Atlas will present their works.

Title: Reconstructing condition-specific regulatory networks across long and short time scales
Speaker: Prof. Sushmita Roy
Abstract:
Changes in gene regulation are hypothesized to play a major role in adaptation and evolution of organism. In recent years, functional genomics approaches have been used to measure different aspects of the gene regulation machinery in single species and extended to multiple species. These functional genomics datasets give us the unique opportunity to develop more comprehensive models of gene regulation. However, doing so requires us to develop novel computational tools that integrate such datasets within one species and across multiple species.

In this talk, I will present computational methods to integrate different types of datasets for the regulatory network of the model organism, Drosophila melanogaster. We show that data integration is key to improved performance and increased coverage of the fly regulatory network.

I will then describe a multi-species analysis framework, which comprises a (1) novel multi-species clustering algorithm, Arboretum that identifies modules in species across large evolutionary distances, and (2) a set of metrics to examine patterns of conservation and divergence in these modules, as well as the factors that drive divergence. We applied our approach to expression profiles measured in 8 species of Ascomycota fungi under glucose depletion and 8 species under heat shock. In both responses, the transcriptomes are captured by five conserved expression modules, however, the degree of gene content conservation in the module was substantially lower in heat shock than glucose depletion, suggesting a stronger conservation of the latter response.

Our approaches for integrating different types of regulatory datasets, within one organism and across multiple organisms, can lead us to systematically understand the structure, function, and evolution of regulatory networks.

Title: A graph-based comparative analysis of three-dimensional organization of chromosomes in yeast and mammals
Speaker: Rob Atlas
Abstract:
Genome-wide maps of chromosomal interactions are becoming increasingly common. Computational tools to analyze these maps and, more importantly, to compare them across contexts are scarce.

In this talk I will describe a novel graph-based clustering approach for detecting sets of interacting genomic loci and comparing them across different tissues and organisms. Our hybrid approach can be applied analogously to 3C data from simple eukaryotes as well as higher eukaryotes such as mouse and humans. We use a penalized cluster quality criteria to determine the number of clusters and develop numerous statistics to systematically examine properties of chromosomal organization.

Application of our approach identifies several properties: (a) the proportion of inter-chromosomal interactions is much higher in yeast compared to mammalian species, (b) in addition to replication forks, distally interacting regions in yeast also exhibit a tendency to be co-regulated (based on gene expression, ChIP- or genetic knockout-based targets of transcription factors), (c) most chromosomal organization is similar between two mammalian tissues, but there are regions that exhibit a tissue-specific interaction pattern, (d) comparison of interaction maps between human and mouse identify extensive conservation of clusters of regions, but we did not find similar conservation between yeast and mammals.

Our graph-based clustering method facilitates a systematic comparison of chromosomal region interaction maps across tissues and species, enabling us to corroborate known findings and identify novel aspects of such interaction maps.