Circulating cellular prognostic biomarkers for cancer
Abstract/Contents
- Abstract
- The peripheral blood contains within it a wealth of information about the underlying biology of cancer patients, yielding insights that can elucidate prognosis. Circulating cellular biomarkers can be examined with a routine and minimally-invasive blood draw, and are a window into physiological and pathological characteristics such as the immune system and tumor metastasis. This thesis will first examine circulating tumor cells (CTCs), and second, the discovery and validation of biomarkers of severe toxicities experienced during immune checkpoint inhibitor therapy. CTCs are a validated cancer biomarker, and their count in the peripheral blood gives critical insights into the patient's condition, enabling disease prognostication and estimation of residual disease. Although they provide rich clinical data, current workflows to enumerate CTCs require highly trained clinicians such as pathologists to manually classify hundreds to thousands of candidate cells in a laborious process that can take up to 30 -- 60 minutes per patient. Here we propose and validate a machine learning pipeline that can greatly reduce the amount of time a pathologist needs to spend classifying images by applying machine learning methods, including transfer learning. We trained several models on lung and renal cancers, and validated them on held-out data from lung, renal and prostate cancers, obtaining area under the receiver operating characteristic curves (AUCs) of 0.94, 0.95 and 0.89 on these cancers respectively for our top performing model. Second, Immune checkpoint inhibitor (ICI) therapy can elicit dramatic response in melanoma, however approximately half of treated patients will develop a severe toxicity leading to acute suffering and often therapy suspension. To this end we interrogated the blood of 13 ICI treated melanoma patients in a high-dimensional single-cell analysis in which we profiled 613,620 cells by CyTOF and 24,807 cells by single cell RNA-sequencing in order to investigate peripheral blood associations with severe toxicity. We found that an activated CD4 memory T cell phenotype was enriched pre-treatment in severe toxicity cases by both CyTOF and scRNA-seq methods, and moreover, we found that patients who went on to develop severe toxicity had higher CD4 T cell clonotype diversity. We investigated this finding further in a cohort of 26 metastatic melanoma patients treated with ICIs for which we obtained pre-treatment blood and applied CIBERSORTx, a method to impute cell phenotype abundances in bulk RNA-seq data. In this cohort we found a significantly enriched immune-phenotype signature pretreatment in patients that go on to develop severe toxicity. Using this as a biomarker training cohort, we combined the immune-phenotype signature with T cell receptor clonotype diversity in an integrative model, yielding an AUC of 0.87. We applied the trained model to a held out 18 patient validation cohort and strikingly, found that the model validated with an AUC of 0.93. We then imputed the immune-phenotype signature in external peripheral blood data across 2 autoimmune diseases in 641 patients compared to healthy controls, and found that it was consistently and significantly associated with autoimmune disease in a meta-analysis. Together, our findings suggest that patients who develop a severe toxicity may have a pre-existing propensity to do so that is unleashed by ICI therapy. Our work establishes a novel pre-treatment toxicity biomarker, and suggests that the cellular composition of pre-treatment blood is critically implicated in toxicity outcomes in ICI-treated melanoma.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2020; ©2020 |
| Publication date | 2020; 2020 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Lozano, Alexander |
|---|---|
| Degree supervisor | Newman, Aaron, (Biomedical data scientist) |
| Degree supervisor | Wang, Shan X |
| Thesis advisor | Newman, Aaron, (Biomedical data scientist) |
| Thesis advisor | Wang, Shan X |
| Thesis advisor | Utz, PJ |
| Degree committee member | Utz, PJ |
| Associated with | Stanford University, Department of Materials Science and Engineering |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Alexander Xavier Lozano. |
|---|---|
| Note | Submitted to the Department of Materials Science and Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2020. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2020 by Alexander Lozano
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...