Approaches for understanding relationships between cancer genomes and clinical cancer features

Abstract

Cancer is a disease of the genome. Understanding what somatic alterations drive cancer and how they affect cancer treatment outcomes is a central goal of cancer genomics. Next-generation sequencing of cancer has led to catalogs of driver alterations in treatment-naive tumors, but in order to understand the evolution of cancer under therapy and develop new treatment strategies, additional studies of cancer samples during treatment are needed.

In order to enable longitudinal studies of advanced cancers on-treatment and examine dynamics of tumor evolution during therapy, we developed an approach for liquid biopsies using cell-free DNA (cfDNA) whole-exome sequencing (WES). We applied ultra-low pass whole genome sequencing (ULP-WGS) to identify somatic copy number alterations in cfDNA and developed a method to estimate cfDNA tumor fraction. By using ULP-WGS to screen blood samples for high tumor fraction, we identified samples with sufficient tumor DNA for cfDNA WES. We compared WES data from matched tumor biopsies and cfDNA samples and found that the detected somatic alterations were concordant. Additionally, we performed longitudinal studies of patients during treatment with targeted therapies and examined the dynamics of resistance. By applying ULP-WGS of cfDNA to large cohorts, we found that approximately one third of patients with metastatic breast or prostate cancer had sufficient tumor fraction for cfDNA WES. Additionally, by studying multi-regional biopsies along with cfDNA WES, we were able to show that cfDNA contained a mixture of tumor clones present in multiple tumors.

Finally, to understand how genomic and transcriptomic features are associated with treatment outcomes for checkpoint blockade immunotherapy (CPB), we performed a meta-analysis of melanoma WES and RNA-Seq data from patients treated with CPB. We combined tumor-intrinsic tumor mutational burden with DNA-based metrics of T or B cell immune infiltration and found that the combination predicted patient survival better than either alone. Additionally, we harmonized RNA-Seq data from multiple cohorts and identified pairs of genes associated with response and survival after immunotherapy, and we validated these associations in independent cohorts. These analyses demonstrate that by combining tumor-intrinsic measurements with immune characterization we can improve predictions of melanoma immunotherapy patient outcomes.