Advisory Committee Chair
Advisory Committee Members
Date of Award
Degree Name by School
Master of Science (MS) College of Arts and Sciences
The study of gene regulatory networks (GRNs) is paramount for continued breakthroughs in understanding the genetic architecture of life and how alterations thereof contribute to disease. The complex spatio-temporal orders within GRNs are studied primarily through the modeling of time-course gene expression data. Recently, gene expression profiling technology has improved greatly in regards to accuracy and resolution, enabling novel methodology development and characterization of GRNs. Many methods have been utilized to model GRNs from gene expression data (correlational measures, time-warping, coherence, and/or causality, etc). In this study, we propose a new approach based on phase locking analysis that compliments existing methods by providing a fundamentally different metric, rooted in statistical physics, for GRN interaction inferences and analysis. Using high temporal resolution gene expression data of Saccharomyces cerevisiae (yeast) during cell cycle, we show that the phase locking metric is a great classifier for interaction and reveals subtle regulatory trends. First, a base-line was established by calculating all phase locking information (for up to 4:3 locking) for every possible gene pairing in the entire yeast genome. Using transcription factor (TF) binding data, we formed positive and negative control groups. We also constructed a high quality positive control group using an in-depth literature compilation and manual annotation by YoungLab at MIT. A sensitivity vs. specificity plot (for correct identification of interaction) was created. The area under the curve (a measure of accuracy) for the 1:1 phase locking index was 0.77 and 0.74 for the alpha factor and CDC28 data sets, respectively. Correlation coefficient yielded an AUC of 0.51. We also applied phase locking analysis to genes co-regulated by cell-cycle TFs and to all 13 3-node subgraphs (found via FANMOD) present within the yeast genome. The results revealed a trend in the locking profile of motif gene pairs, a trend in the locking as a function of number of edges for subgraphs, and a potential increase in multi-state stability for high edge number subgraphs. Our research has led to the conclusion that phase locking analysis shows great promise for GRN modeling and should be further explored by the systems biology community.
Roberts, Herbert Keith, "Characterization of Emergent Spatial-Temporal Orders in Transcription Regulatory Networks using Phase Locking Analysis" (2013). All ETDs from UAB. 2841.