Boston University
Spatial omics tools for the characterization of triple negative breast cancer heterogeneity
Abstract
dc:description.abstractTriple negative breast cancer (TNBC) is the most aggressive breast cancer subtype and has the lowest survival rates. Contributing to these poor outcomes are the limited targeted therapies available when compared to other breast cancer subtypes, and the paucity of reliable predictors to treatment response. Emerging evidence starts to indicate that the behavior of TNBC is not only shaped by the cancer cells themselves but also by the non-cancer cells and how they are spatially organized within the tumor microenvironment (TME). Indeed, cells such as immune, stromal, and vascular cells are organized into recurrent multicellular niches within the TME, and rather than the cellular abundances alone, these multicellular spatial topographies carry prognostic weight for survival outcomes. Spatial omics technologies can now resolve this organization directly by measuring molecular analytes (e.g. RNA or protein) within intact tissues, allowing researchers to resolve cell types and understand biological processes that are spatially organized. However, while the development of turnkey platforms have greatly enhanced the ability to generate these complex datasets, the limiting factor has become analysis rather than data generation. To tackle this bottleneck a growing set of platforms, that vary in both modality and spatial resolution, have been developed. Yet, most spatial data analysis tools are vendor- or technology-specific and lack the tools to represent or integrate one or more spatial datasets, each with their own unique raw data structures, resolutions and modalities, within one coherent framework. At the same time, the latest emerging whole-transcriptome platforms generate data at a scale that outpaces existing analytical infrastructure. This dissertation jointly addresses both the biological and computational gaps thereby ensuring that new computational methodologies are grounded in biological utility. Specifically, we built a scalable, technology-agnostic spatial-analysis infrastructure within the R ecosystem and then applied it to TNBC, showing how it leads to accelerated new biological insights. First, we present Giotto Suite, a new framework underlying a modular analysis toolbox that can represent and integrate spatial multi-omics data across virtually all molecular modalities and spatial scales. Notably, Giotto Suite is directly interoperable with other large genomics and geospatial software ecosystems, further extending its impact and utility. We then developed an innovative on-disk extension built for atlas- and whole-slide-scale data. By combining lazy columnar storage and query (GiottoDisk) together with a spatially tiled processing framework (tilework), we can store and analyze spatial omics data in a flexible and very fast manner, respectively. Furthermore, by creating a static and server-free interactive viewer (GiottoLens), users of Giotto Suite can view and manually annotate these very large datasets. These tools were then applied to characterize TNBC spatial heterogeneity across a multi-patient cohort that was profiled by single-cell-resolution spatial transcriptomics. After initial fast data processing and cell-type annotation, recurrent multicellular niches were identified through a community detection algorithm over spatial graphs. These niches were then further characterized through spatial autocorrelation of cellular organization, allowing the architecture of the TME to be systematically compared across patients. As such we demonstrated how our newly developed framework works in practice, thereby surfacing new features of inter-patient spatial heterogeneity in TNBC. Finally, our software engineering and data analysis work provides an accessible, comprehensive, and statistically grounded foundation for spatial data analysis and visualization at the scale that current and forthcoming platforms demand.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Genetics & Genomics
- Grantor
- Boston University
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Chen, Jiaji George
- Advisor dc:contributor.advisor
-
- Dries, Ruben
- ORCID dc:identifier.orcid
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53553