Functional Analysis of Transcriptional Response During Impaired Cutaneous Wound Healing
Tuesday, August 15, 2017
10:00 AM-12:00 PM
BIOMED PhD Thesis Defense
Functional Analysis of Transcriptional Response During Impaired Cutaneous Wound Healing in a Murine Model of Diabetes
Sina Nassiri, PhD Candidate, School of Biomedical Engineering, Science and Health Systems, Drexel University
Kambiz Pourrezaei, PhD, Professor, School of Biomedical Engineering, Science and Health Systems, Drexel University
Issa Zakeri, PhD, Professor, Dornsife School of Public Health, Drexel University
Whole-genome transcriptional profiling holds great potential to enhance our understanding of the pathophysiology of diabetic wound healing. However, analyzing time series transcriptional data is challenging due to small sample size, irregularly spaced time points, missing values, noisy observations, and high-dimensionality of the data. Additionally, the temporal order and potential dependence of measurements may introduce complex correlation structure in the data, which if ignored can lead to loss of information and reduced statistical power. A natural and fruitful approach to circumvent these challenges and simultaneously take the temporal order and potential dependence of observations into account is to postulate transcriptional response as a smooth function over time. Discrete observations can then be viewed as noisy realizations of this smooth function. This idea is at the core of a rapidly developing branch of statistics known as functional data analysis (FDA). The overarching goal of this thesis was to devise and implement a comprehensive functional approach to analyze time series transcriptional data on impaired cutaneous wound healing in a murine model of diabetes. We pursued this goal via the following specific aims.
In Aim-1, we investigated gene expression dynamic from a single-gene perspective. A fundamental objective of analyzing transcriptional response during wound healing is to identify genes that exhibit change of expression over time or have distinct temporal pattern between two or more study groups. To this end, we compared the expression profile of each gene over the course of healing in diabetic mice (db/db) versus controls (db/+). Moreover, we demonstrated the unique utility of FDA in pointwise comparison of gene expression profiles at time points beyond those dictated by study design.
In Aim-2, we investigated gene expression dynamic from a gene set perspective. Extracting actionable information from genome-wide dynamic regulation of gene expression is not straightforward. Gene set analysis is one approach to decipher the vast pool of information obtained from high throughput transcriptome data by utilizing prior knowledge about the functional role and biological association of genes. In the second aim, we utilized FDA to extend an existing procedure originally developed for static studies with simple case-control designs, to time course studies with complex scenarios. Furthermore, we proposed an intuitive functional test to assess differential expression of a given gene set over the course of healing in db/db wounds compared to db/+.
In Aim-3, we investigated gene expression dynamic from a system-level perspective. It has become increasingly evident that cellular and molecular components of biological systems do not act in solitary. Instead, they interact as elements of highly interconnected networks. In the third and final aim, we investigated the network architecture of gene coexpression during impaired diabetic wound healing. Taking advantage of smooth curve representation of time series gene expression data, we used a functional measure of pairwise dynamic coexpression to construct gene coexpression networks for db/db and db/+ wounds. Subsequently, we characterized the structural properties of each network independently and in comparison to one another.
Taken together, by building upon the rich and flexible framework of FDA we circumvented various challenges associated with analyzing time series data, and demonstrated its utility in different aspects of gene expression analysis, from differential expression, to gene set testing, to network analysis. Collectively, this work allows us to gain better understanding of cellular and molecular mediators of impaired healing, and sheds light on candidate genes and signaling pathways that are implicated in impaired diabetic wound healing.