PhD Proposal: Studying tumor evolution with emerging sequencing technology
Yuelin Liu
Thursday, August 24, 2023, 10:00 am-12:00 pm 
Abstract
Cancer is a heterogeneous disease: tumors not only present varied characteristics across cancer types and different patients, tumors within the same patient oftentimes contain heterogeneous sub-populations, as well. In the series of work discussed below, we focus on the latter. A better understanding of such intra-tumor heterogeneity may lead to more accurate cancer diagnostics, prognostics, and more effective and targeted treatments. Since its proposal in the 1970s, the clonal evolution theory of carcinogenesis has provided a model with which we can study how intra-tumor heterogeneity arises. The accumulation of genomic and epigenomic alterations over the course of tumor evolution can be studied, and key alteration events can be identified using methods commonly employed in species evolution, such as phylogeny reconstruction. Under this analytical framework, given sequencing data from a tumor sample, a tumor phylogeny can be constructed based on the observed somatic alteration events, and the evolutionary trajectory or the tumor can be inferred from the constructed phylogeny. There has been a vast body of studies that leverage mutations and copy number alteration information from bulk and/or single-cell sequencing data generated from next-generation short-read sequencing technologies. However, the emergence of new multi-omic single-cell sequencing, spatially-resolved RNA sequencing, and third-generating long-read sequencing technologies has brought both new opportunities and challenges to the study of tumor evolution.
In the first piece of work, we leverage single-cell sequencing data sets that have matching CpG methylation and RNA sequencing information in tumor phylogeny reconstruction. We introduce the first distance-based computational method, Sgootr, for inferring a tumor's single-cell methylation-based phylogeny and jointly identifying lineage-informative CpG sites which harbor changes in methylation status that are retained along the lineage. We apply Sgootr on single-cell bisulfite-treated whole genome sequencing data of multi-regionally-sampled tumor cells from 9 metastatic colorectal cancer patients, as well as multi-regionally-sampled single-cell reduced-representation bisulfite sequencing data from a glioblastoma patient, incorporating copy number information called from the matching scRNA-seq data whenever possible. We demonstrate that the tumor phylogenies constructed reveal a simple model underlying tumor progression and metastatic seeding.
The two pieces of proposed work we discuss next are parts of a series of studies on melanoma. Beginning in melanin-producing cells called melanocytes, melanoma is the most serious form of skin cancer. Pan-cancer analysis on exome and RNA sequencing data from The Cancer Genome Atlas and other cohorts shows that melanoma is one of the most heterogeneous cancers. A better understanding of the key genomic and epigenomic events that characterize the diverse subclonal populations in melanoma may reveal key insights into what drives its progression and therapeutic resistance. The data generated for the studies stem from the mouse B2905 melanoma cell line. In the first proposed work, 24 distinct clonal sublines were derived in vitro from single cells of the cell line culture, and the genetically homogeneous population from each subline was subject to Oxford Nanopore Technology long-read sequencing. Empowered by the possibility to directly call 5mC base modifications and to perform haplotype phasing with Nanopore long-read data, we will develop a novel computational pipeline to integrate insights in detected CpG methylation, small variants, and structural variants in our study of melanoma evolution. In the second proposed work, mice were injected with cells from the cell line culture and assigned to control or anti-CTLA-4 immune checkpoint blockade therapy group. The tumors were then resected at different time points after injection and treatment, and tissue slices were subject to 10x Visium spatially resolved RNA sequencing. We derive subclone-specific expression signatures from tumor phylogenies constructed from scRNA-seq data of similarly obtained tumors, and use them to map subclones onto spatial coordinates through deconvolution. This allows a closer examination of subclone growth dynamics and interactions with immune cells under different treatment conditions.
In the first piece of work, we leverage single-cell sequencing data sets that have matching CpG methylation and RNA sequencing information in tumor phylogeny reconstruction. We introduce the first distance-based computational method, Sgootr, for inferring a tumor's single-cell methylation-based phylogeny and jointly identifying lineage-informative CpG sites which harbor changes in methylation status that are retained along the lineage. We apply Sgootr on single-cell bisulfite-treated whole genome sequencing data of multi-regionally-sampled tumor cells from 9 metastatic colorectal cancer patients, as well as multi-regionally-sampled single-cell reduced-representation bisulfite sequencing data from a glioblastoma patient, incorporating copy number information called from the matching scRNA-seq data whenever possible. We demonstrate that the tumor phylogenies constructed reveal a simple model underlying tumor progression and metastatic seeding.
The two pieces of proposed work we discuss next are parts of a series of studies on melanoma. Beginning in melanin-producing cells called melanocytes, melanoma is the most serious form of skin cancer. Pan-cancer analysis on exome and RNA sequencing data from The Cancer Genome Atlas and other cohorts shows that melanoma is one of the most heterogeneous cancers. A better understanding of the key genomic and epigenomic events that characterize the diverse subclonal populations in melanoma may reveal key insights into what drives its progression and therapeutic resistance. The data generated for the studies stem from the mouse B2905 melanoma cell line. In the first proposed work, 24 distinct clonal sublines were derived in vitro from single cells of the cell line culture, and the genetically homogeneous population from each subline was subject to Oxford Nanopore Technology long-read sequencing. Empowered by the possibility to directly call 5mC base modifications and to perform haplotype phasing with Nanopore long-read data, we will develop a novel computational pipeline to integrate insights in detected CpG methylation, small variants, and structural variants in our study of melanoma evolution. In the second proposed work, mice were injected with cells from the cell line culture and assigned to control or anti-CTLA-4 immune checkpoint blockade therapy group. The tumors were then resected at different time points after injection and treatment, and tissue slices were subject to 10x Visium spatially resolved RNA sequencing. We derive subclone-specific expression signatures from tumor phylogenies constructed from scRNA-seq data of similarly obtained tumors, and use them to map subclones onto spatial coordinates through deconvolution. This allows a closer examination of subclone growth dynamics and interactions with immune cells under different treatment conditions.
Examining Committee
Bio
Yuelin Liu is a PhD student in Computer Science at University of Maryland, College Park (UMD), and a predoctoral visiting fellow at the National Cancer Institute (NCI/NIH). Her research is in computational cancer genomics with a focus in tumor evolution. She is advised by Dr. Cenk Sahinalp at NCI and Prof. Mihai Pop at UMD.
This talk is organized by Tom Hurst