This study investigates Mandarin discourse markers from both functional and prosodic perspectives. Discourse markers are defined as sequentially dependent elements which bracket units of talk (Schiffrin 1987). In this study, I focus on three discourse markers, ranhou ‘then’, wo juede ‘I think/feel’, and meiyou ‘no, not’. Using video-taped everyday conversation, I examine their functional categories and prosodic features, including duration, pitch range, and stress.

My qualitative and quantitative analyses show that, first, the discourse functions of these markers are prosodically stronger than their lexical functions. Newly developed discourse functions, such as the topic-shifting function of ranhou and floor-claiming function of meiyou, take on a special prosodic design that distinguishes them from other functions.

Second, different prosodic features can have distinct functional and interactional relevance and thus prosody should not be treated as a unitary dimension in conversation. Discourse markers can have incongruent prosodic features, with one prosodic dimension strong and other dimension(s) weak. The ostensible ‘inconsistency’ is in fact designed so as to fulfil particular interactional needs.

Lastly, given the prosodic difference between lexical and discourse functions discussed, this study supports the view that the development of discourse markers should be considered an independent diachronic process, i.e. pragmaticalization, rather than a subtype of grammaticalization.

Additionally, the present study has methodological contributions in that it has demonstrated the feasibility of integrating a quantitative approach and statistical tests into the study of discourse markers, which are traditionally studied qualitatively. As a complement to the qualitative approach, quantitative methods are able to provide a more objective angle to evaluate the relationships between prosody and function.

Live-Cell imaging and other fluorescence-based measurement technologies provide the best time resolution of a few minutes to seconds. However, despite recent advances with multiplexing reporters the number of unique reporters that can be recorded in a Cell is limited to a few dozen. Most studies still measure only one or two reporters at the same time at high temporal resolution. Unlike Live-Cell imaging, Single-cell RNA-seq and related technologies provide a genome-wide view for every time-point. For Single-cell RNA-seq however, the best time resolution is 30 minutes at best, while 1 hour is typical for studying fast dynamics. Therefore, to provide both high temporal resolution and genome-wide measurement of single-cell RNA we have invented Chrono-seq. Chrono-seq can take samples as fast as six minutes and as long as arbitrarily needed.

Motivation

Results

Availability

I have been working on two projects through my master thesis. The first project described in Chapter 1 is focusing on studying the regulatory mechanisms of specific methylated DNA motifs using optimized CAPTURE technique. DNA motifs are the sequences that are presumed to have biological functions in gene regulatory, however, the regulatory compositions of most motifs remain unknown. CAPTURE is a perfect tool that we can optimize and apply on our interest methylated DNA motifs.

Epigenetic mechanisms of gene regulation such as DNA methylation and chromatin modification are playing an important role in the development of pathology have been demonstrated, but the epigenetic regulation of gout disease is unclear. The second project in Chapter 2, we modified and applied Atac-seq and RNA-seq to study the difference of open chromatin pattern and gene expression respectively between gout patients and healthy individuals.

DNA methylation plays crucial roles in many biological processes and abnormal DNA methylation patterns are often observed in diseases. Recent studies have shed light on cis-acting DNA elements that regulate locus-specific DNA methylation. More importantly, these new discoveries have shown potentials in clinical application.

In this thesis, I first interrogate the current biological foundation for the cis-acting genetic code that regulates DNA methylation. This process involves transcription factors, histone modifications, and DNA secondary structure. In chapter 2, we demonstrate how to find the functional motifs that regulate DNA methylation. We have analyzed 34 diverse whole-genome bisulfite sequencing datasets and have identified 313 identified motifs, including 92 and 221 associated with methylation (methylation motifs, MMs) and unmethylation (unmethylation motifs, UMs), respectively. We show that these motifs are associated with local methylation level, and motif disruption of by mutation leads to significantly altered methylation level of the CpGs in the neighbor regions. Combined with somatic mutations, these motifs improve the prediction of cancer subtypes and patient survival.

DNA motif analysis frequently requires intuitive understanding and convenient representation of motifs. In chapter 3, I review how the motifs are typically represented as position weight matrices (PWMs) and propose a new wildcard-style consensus sequence representation based on mutual information theory and Jenson-Shannon Divergence. We name this representation as sequence Motto and have implemented an efficient algorithm with flexible options for converting motif PWMs into Motto from nucleotides, amino acids, and customized alphabets. On the other hand, experimental validation of cis-acting DNA elements benefits from the recent advancement of CRISPR/Cas9 mediated genetic screening. In chapter 4, I present CRISPY, a lightweight, robust CRISPR screening pipeline that unifies single-sgRNA and CREST-seq screening protocols and is capable of profiling peak candidates with existing data of histone modifications, DHS, and ATAC-seq in human and mouse.

Combined together, our studies have provided new insights on how genetic code regulates DNA methylation and can be applied to clinical applications. In addition, we provide the tools to efficiently represent the motifs and evaluate their functions in a high-throughput manner.

In the field of functional genomics, understanding the 1D sequence elements and 3D interactions of chromatin is of utmost importance. Previous studies have identified hubs as crucial nodes in the 3D chromatin landscape, and demonstrated their essentiality even in quiescent regions without any known 1D epigenetic annotations. These findings have shed light on the interplay between the linear genome and 3D contacts in cellular functions. However, a comprehensive comparison between hubs and a clear mechanism explaining their phenotypic consequences are still lacking.

In this dissertation, our primary objective was to address these gaps in knowledge. Initially, we utilized machine learning to guide a CRISPR screening of hubs, resulting in an improved candidate pool and a significantly higher discovery rate. Next, we employed statistical tests and machine learning algorithms to classify hub essentiality, revealing no discernible genetic or epigenetic differences between essential and nonessential hubs. However, we observed distinct spatial contacts and opposing effects on cell growth for neighboring hubs in the linear genome. Subsequently, we selected a pair of such hubs located in an epigenetically quiescent region and conducted in-depth analyses using single-cell experiments.

Our investigations validated the differential impacts of these hubs on gene expression, chromatin accessibility, and chromatin organization. Furthermore, we discovered that deletion of the essential hub resulted in alterations in genetic network activity and increased Shannon entropy of chromatin accessibility, exceeding the effects caused by deletion of the nonessential hub. These findings suggest the critical role of hubs in maintaining an orderly chromatin structure. By providing novel insights into the systemic functions of non-coding regions in the human genome, our study enhances our understanding of gene regulation at a broader scale.