Analysis of DNA deformation patterns in nucleosome core particles and identification of gene regulatory regions based on dimensionality reduction and continuous wavelet transform

Abstract

The superhelix structure of a nucleosome formed by DNA wrapping around the histone octamer is important in enabling the DNA to finally fold into chromatin, and directly influences DNA replication, recombination, repair and transcription. The bending of DNA in a nucleosome has been proven to be very anisotropic and influenced by many factors. To establish the relationship between a nucleosome positioning signal and the geometric characteristics of the constituent DNA fragments is helpful for the understanding of the genetic regulation mechanism. On the other hand, it is a basic task of genome annotation to locate genes and their regulatory regions along a DNA sequence. A promoter, as a kind of important DNA transcription regulatory unit, is located near a gene and contains critical response elements that can be identified and bound by RNA polymerase and transcription factors to initiate gene transcription. Computational methods of promoter prediction mostly rely on the conserved cis-acting sequence motifs, such as TATA-boxes, CpG islands and CAAT boxes. However, as proposed recently, structural properties of DNA are equally important in describing the characteristics of gene and gene regulatory regions, of which the new prediction methods can take advantage. This thesis concentrates on two aspects: (1) the analysis of DNA deformation patterns in nucleosome core particles based on dimensionality reduction and continuous wavelet transform, and (2) the characterization and prediction of eukaryotic promoters by applying dimensionality reduction and support vector machine (SVM) on multiple DNA physicochemical properties. Dimensionality reduction algorithms, principal component analysis (PCA) and isometric feature mapping (Isomap), are used to derive structural signals from the DNA structural data of nucleosome crystals. The analysis of this structural signal by continuous wavelet transform has successfully revealed the common regularity of nucleosome DNA deformation and also the peculiarity of structural configurations in nucleosomes with a particular histone or DNA composition, or special ligands. The relationship between the constituent dinucleotides and the signal energy distribution shows that the fluctuation of a structural signal is sensitive to certain dinucleotide types. We also investigate and compare the features in the structural profiles of core promoter regions in several typical eukaryotes. Promoters are aligned at the transcription start site (TSS) for each eukaryote type and we obtain an averaged promoter representative for this eukaryote type. Then the nonlinear dimensionality reduction Isomap algorithm is applied on the averaged promoter model, which is described by a set of physicochemical parameters, to separate the comprehensive structural profile. The result shows that eukaryotic core promoters have very special structural characteristics, which are independent of their DNA sequence content. Besides, a new eukaryotic promoter prediction method is proposed by integrating the physicochemical profile of the DNA sequence and the SVM. The application of this promoter prediction model on the genomic data of the Drosophila melanogaster shows that the new model obtains a good performance on specificity and sensitivity.

Date of Award	15 Feb 2013
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Hong YAN (Supervisor)