Computational Research
Project 1: Computational Analysis Of Differentiation In Cancer Progression
Sylvia K. Plevritis, PhD (Lead)
Daphne Koller, PhD (co-Lead)
David L. Dill, PhD (co-Lead)
We will develop and apply a variety of synergistic computational approaches to the problem of understanding cellular differentiation and self-renewal programs in our experimental studies of FL, AML, and mouse T-ALL. Our aims are split into three groups addressing key biological questions: 1) reconstructing molecular regulatory networks; 2) inferring progression processes in cancer; 3) developing Boolean and executable models of cellular processes. When integrated, these methods will permit us to computationally model the regulatory influences underlying cancer progression, particularly relating to the role of differentiation, which is the overall theme of our CCSB application.
Koller Project: Develop computational methods for reconstruction of regulatory networks underlying differentiation in normal and malignant cells
- Learn regulatory networks with additional molecular regulators. We will extend our previously developed computational approaches for regulatory network reconstruction to integrate new types of data generated by Projects 2-4
- Exploit background knowledge in learning gene regulatory networks. Prior information, such as known interactions, will be incorporated into our algorithms to improve their predictiveness.
- Learn models of shared and differential regulation across different samples. We will apply methods from transfer learning to incorporate knowledge of cellular hierarchies learned from experimental Projects 2-4, to learn networks of shared and differential regulation between subpopulations of tumor cells.
Plevritis Project: Develop and validate computational methods for inferring cancer progression
- Recover an ordering of experimental samples by identifying and tracking a gradual change in global molecular expression data. We will employ concepts from graph theory to infer an ordering across experimental samples. We will apply these methods to order normal versus cancer samples using mRNA and DNA array data and test for underlying signatures of differentiation. We will also perform this analysis to order single cells of normal versus cancer from flow cytometry data and test the presence of a cellular hierarchy defined by stages of differentiation.
- Apply principles of topology to reveal patterns of cancer progression from global molecular profiling data. We will employ our topological methods of shape recognition, Mapper and Persistent Homology, to identify patterns of progression in normal versus cancer. We will extend this idea to compare the topology of regulatory networks of normal versus cancer and to identify how the connectedness of related sub-networks is perturbed.
- Develop a multi-scale model of cancer that estimates cellular level dynamics from tumor growth kinetics. We will develop a hybrid model that describes the differentiation hierarchy of cells in normal and malignant populations.
Dill Project: Develop and validate Boolean methods for system modeling and data analysis.
- Find causal relationships between regulators and their downstream targets using Boolean implications. Logical relationships between genes will be derived from data extracted from large publicly available data sets
- Predict patterns of expression of developmentally regulated genes using Boolean implications. We will utilize Boolean implications derived in Aim 1 to refine identification of key regulatory genes involved in differentiation and self-renewal in FL, AML, and mouse models
- Develop and validate executable in-silico models for cell signaling related to AML, T-ALL, and lymphoma, based on an extended version of the open source Pathway Logic Viewer. We will develop a queryable model of signaling pathways, initially for the B-cell receptor signaling pathway that is central to investigations in Project 3.
