Contact Principal Investigator / Project Leader: Yang, Mary Organization: University of Arkansas Little Rock Title: DEVELOP NOVEL DEEP LEARNING AND COMBINATORIAL OPTIMIZATION METHODS TO IDENTIFY KEY DISEASE REGULATORY ELEMENTS FOR SINGLE-CELL DATA Abstract Text: Description: Traditional bulk sequencing measures the average of cell population constituents, inevitably masking the intrinsic cell-to-cell heterogeneity. Single-cell technologies, on the other hand, enable a high-resolution measurement for each individual cell, providing new opportunities to capture cell population diversity and dissect the heterogeneity of complex diseases. Meanwhile, the high-sparsity and the relatively small number of sequencing reads pose new data analytic challenges. In this proposed project, we will develop innovative computational methods for single-cell RNA sequencing (scRNA-seq) data analysis and integration to identify key regulatory elements that underlie disease heterogeneity and drive disease development. The scRNA-seq data contains substantial proportion of zero expression counts due to low capture efficiency and stochastic gene expression. We will develop a novel data-driven deep learning model to recover the missing values. Our model utilizes a deep learning algorithm to capture complex and latent distributions of missing values without assuming an underlying distribution, thus, ensuring effective performance across various scRNA-seq generated by different protocols. scRNA-seq profiles enable characterization of unique transcriptome for each cell type. We hypothesize that disrupted expression patterns accompanying the disease development in different cell types are controlled by sequential alterations of the activity and connectivity in the regulatory networks. Hence, using scRNA-seq data, we will first infer cell lineage trajectories. Then, we will develop a novel deep neural network method to reconstruct cellular regulatory networks according to pseudo-time ordering of the cell types. With a new network alignment model, we will exploit the dynamic changes of regulations in the disease process, revealing key regulators and providing cell type-specific drug targets. The fulfillment of the proposed project will facilitate single-cell genomic and biomedical research efforts allowing for a much broader, cross-disciplinary understanding of the underlying mechanisms of complex diseases. The proposed project will be devised into capstone projects and will be primarily completed by undergraduate students under the PI's supervision with the assistance of a graduate student. The project will serve as a vehicle to equip undergraduate students with essential research skills and interdisciplinary knowledge, and to stimulate the students' ambition to pursue careers in the biomedical science. This project will create a multidisciplinary platform in a comprehensive university setting that encourages undergraduate students to engage in biomedical research.!
This R15 project aims to provide novel deep learning and hybrid statistical methods to tackle challenges with a focus on single-cell genomic data to improve our understanding of disease evolution mechanisms and advance treatment strategies for the complex diseases. Our approaches enable the reconstruction of cell type-specific regulatory networks and infer dysregulated pathways in different cell types and the integration of the research efforts made on the bulk-level, hence, lead to further discoveries, accurate diagnosis and new cell type-specific therapies. With the processes of this R15 project, we will engage and continue to recruit undergraduate students from diverse backgrounds focusing on their research skills development and incorporate the scientific investigations for this project into their college education and provide broader careers.
