With the advent of modern-day sequencing, imaging, and spatial proteomics technologies, the rate at which data is being generated in the oncology clinic and at the bench is far surpassing the rate at which the computational resources to process this deluge of data are improving. There exist petabytes of openly available cancer pathology imaging data, both from traditional H&E staining and novel spatially resolved proteomics methods such as CODEX, that contain information that is highly relevant for cancer treatment and therapeutic discovery. In order to leverage this data to improve cancer treatment and drug discovery initiatives, algorithms and machine learning methods that enable it to be efficie...

Read More

With the advent of modern-day sequencing, imaging, and spatial proteomics technologies, the rate at which data is being generated in the oncology clinic and at the bench is far surpassing the rate at which the computational resources to process this deluge of data are improving. There exist petabytes of openly available cancer pathology imaging data, both from traditional H&E staining and novel spatially resolved proteomics methods such as CODEX, that contain information that is highly relevant for cancer treatment and therapeutic discovery. In order to leverage this data to improve cancer treatment and drug discovery initiatives, algorithms and machine learning methods that enable it to be efficiently compressed must be designed and implemented. My work has focused on making large scale computing efficient for applications related to bioinformatics and genomics, development of computational methods for microbiome analysis that allow for mechanistic understanding of interactions between individual microbes and the host, development of machine learning enabled liquid biopsy techniques that track CD8+ T cell lineages in the context of immunotherapy, and implementation of algorithms that will lead to search technologies that can contextualize petabytes of cancer imaging data. Leveraging the expertise gained from my training, my group will have three main priorities: 1) building a contextual search engine for clinical cancer imaging data (Research Aim 1), 2) building extensible deep learning tooling for oncologists to leverage domain expertise without substantial computational background (Research Aim 1), 3) and discovering metagenomic gene modules contributing to effectiveness of cancer immunotherapy and validating these with spatially resolved proteomics approaches (Research Aim 2).

Read Less