UTHealth Houston - Graduate School of Biomedical Sciences

Genomic and Immunologic Characterization of Appendiceal Adenocarcinoma: Methods for Studying Rare Cancers with Limited Data

Matthew James Lastrapes, BSE (Advisor: Paul Scheet, PhD)

Appendiceal adenocarcinoma (AA) is a rare disease, affecting only a few thousand individuals each year in the United States. However, incidence is rising, particularly among young adults. There are no disease-specific therapies for AA because its biology remains poorly understood, in part due to limited sample sizes but also the nature of data. Appendiceal cancers tend to contain large amounts of extracellular mucin, and therefore exhibit low cellularity that dilutes tumor-specific signals in research data. Existing studies of AA have relied largely on targeted gene panels, which limit genome-wide analyses. This dissertation addresses these limitations, and develops and applies novel methods and technologies to AA data to extract meaningful biological insight from two angles: somatic genomics and the tumor microenvironment.

First, we evaluated the sensitivity of hapLOHseq, a method for allelic imbalance detection, to degradation of its input data by downsampling the read counts at genotyped sites and introducing errors into statistically estimated haplotypes. Using tumors from The Cancer Genome Atlas (TCGA) with matched whole exome sequencing (WES) and SNP microarray data, we tested allelic imbalance event calls from WES using array events as truth through a range of coverages and accuracies, identifying that AI detection remained robust at coverages above 10X and phasing accuracies above 85%. Since clinical coverage is typically well above 10X, we concluded that phasing accuracy is the more likely limiting factor in AI detection from clinical data using hapLOHseq.

Second, we developed pairPHASE, a novel haplotype phasing method intended for limited datasets, to address this need. pairPHASE uses pairwise linkage disequilibrium estimates computed from an external reference population to phase individual samples without needing cohort-level information. When evaluated on WES and RNAseq data from three TCGA cohorts, pairPHASE produced haplotypes with higher switch accuracies than MaCH, an established HMM-based phasing method. Further, AI detection using these haplotypes was comparable to using those generated by MaCH, illustrating the downstream utility of pairPHASE for clinical datasets.

Third, we extended AI detection analysis to an internal dataset of 98 AA patients sequenced using WES. When compared to a standard somatic copy number alteration detection method, CNVkit, the combined pairPHASE + hapLOHseq pipeline showed higher detection rate of more subtle events, either due to limited tumor purity or low clonality. Additionally, hapLOHseq was able to detect copy-neutral loss of heterozygosity (cnLOH) events invisible to CNVkit. Imbalance events detected by hapLOHseq were supported by analysis of matched RNAseq data. hapLOHseq-based AI burden was significantly associated with TP53 mutation status, as has been described in previous studies of other cancers.

Finally, we characterized the immune microenvironment of 9 AA pre-treatment tumors from a phase II clinical trial of combined atezolizumab and bevacizumab using spatial proteomics. Identification and classification of distinct immune aggregates suggested mature tertiary lymphoid structures (TLS) as potential biomarkers of improved outcomes in AA patients.

These projects provide meaningful insight for the analyses of genomic and immunologic data under the constraints introduced by a rare cancer like AA, by providing novel methods and applying cutting-edge technologies in genomics and immunology. This framework can be extended to future studies of AA, or adapted to study other rare cancers or diseases in general. 

Advisory Committee:

• Paul Scheet, PhD, Chair
• Veronica Fedirko, PhD
• Humam Kadara, PhD
• John Paul Shen, MD
• Jennifer Wargo, MD

Join via Zoom (Please contact Mr. Lastrapes for the details of his Zoom meeting information.)