Skip to Main Content
An official website of the United States government
Login Change: CDAS is now using the NIH Researcher Auth Service (RAS) to manage our existing login options. When logging into CDAS, you will be taken to RAS, where you can select your preferred login provider to access CDAS.

Quantifiable TCR repertoire changes in pre-diagnostic blood specimens among high-grade ovarian cancer patients

Principal Investigator

Name
Bo Li

Degrees
PhD

Institution
Children's Hospital of Philadelphia

Position Title
Associate Professor

Email
lib3@chop.edu

About this CDAS Project

Study
PLCO (Learn more about this study)

Project ID
2025-0079

Initial CDAS Request Approval
Sep 10, 2026

Title
Quantifiable TCR repertoire changes in pre-diagnostic blood specimens among high-grade ovarian cancer patients

Summary
High-grade serous ovarian carcinomas (HGSOC) comprise 70% of sporadic ovarian cancers (OC) and caused most OC-related deaths. HGSOCs are dominantly diagnosed at stage III or IV, with 5-year survival only 30%. If diagnosed at stage-I when tumors are locally confined, the disease is curable with over 90% 5-year survival. However, majority of HGSOCs originated from microscopic tumor lesions called Serous Tubal Intraepithelial Carcinoma, or STIC, in the fallopian tube, which subsequently spread to the ovaries and peritoneal cavity. This property of STIC renders HGSOCs to be metastatic at very early stage of tumor progression and only less than 10% HGSOC patients are diagnosed at stage-I. Existing biomarkers, or imaging scans, such as serum CA-125, HE-4, transvaginal ultrasound, etc, suffer from low sensitivity to detect the small HGSOC lesions at early-stage. As a result, large screening trials have shown that these clinical modalities cannot reduce OC-related mortality. On the other hand, HGSOCs frequently develop p53 mutations and deficiencies in DNA mismatch repair machinery during early progression. These events create tumor antigens that can trigger adaptive immune responses. As a critical component in the immune system, the T lymphocytes specifically target these antigens via their T cell receptors (TCR), followed by T cell activation and massive proliferation. Finding cancer-associated changes in the blood T cell repertoire thus provides a rational solution to the diagnosis of HGSOCs in its early stage, which usually happens years before their conventional diagnosis. Here, we propose to develop and validate a TCR-based non-invasive biomarker for the prediagnostic detection of HGSOCs. The search for cancer-related signals among T cells is challenging given the enormous diversity of the human TCR repertoire and limited knowledge of T cell antigen specificity. My team pioneered this direction by developing computational methods to study the TCR repertoire, including TRUST to assemble TCRs from cancer genomics datasets, iSMART to group antigen-specific TCRs, DeepCAT to perform de novo prediction of cancer-associated TCRs and GIANA for ultrafast TCR clustering and disease classification. To prepare for this task, we have collected TCR repertoire sequencing samples from patients with malignant or benign ovarian malignancies and Nurses’ Health Study participants with blood specimens collected up to 5 years before HGSOC diagnosis. This unique sample cohort allowed us to investigate the landscape of blood TCR repertoire in OC patients. Using a new TCR embedding method, we identified significantly altered repertoire components in HGSOC compared to benign patients. Importantly, these components captured a quantifiable TCR repertoire change up to 4 years before conventional HGSOC diagnosis. Motivated by this exciting result, we will leverage the PLCO patients with prediganostic samples collected within 6 years prior to cancer diagnosis (n=81) to develop a TCR-based biomarker through a new computational framework (Aim 1) and perform independent validation using the PLCO samples (Aim 2), with the goal of reaching a positive prediction value of 5% for early-stage HGSOCs.

Aims

In this proposal, we will take advantage of our existing ovarian sample cohort and the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer screening trial, to develop and validate a novel TCR-based assay. PLCO includes >50 HGSOC patients with serial blood samples collected up to 6 years before diagnosis, covering the most variable window (-4 to -2 years) of the TCR repertoire. Due to the rarity of stage-I HGSOCs, direct investigation of the immunological features of these patients to develop an early-stage biomarker is nearly impossible. Therefore, these prospectively collected, prediagnostic PLCO samples will provide an ideal, and almost exclusive opportunity for the independent assessment of this TCR-based biomarker:
Aim 1. TCR assay development and evaluation. Our preliminary analysis of the TCR-seq samples achieved a moderate accuracy (AUC=0.79) and implicated the path towards better performance by increasing training sample size and including more information of TCR/antigen interaction. In this Aim, we will design a new statistical method to boost the performance through: 1) inclusion of pan-cancer TCR-seq samples in the training data to learn a comprehensive adaptive immune roadmap for early-stage tumors; 2) development of a fine-mapping of the TCR numeric embedding space to search for cancer-associated TCR motifs; 3) incorporation of patient HLA genotype information to precisely search for informative TCR clusters. Quantifiable TCR repertoire components will be identified using pan-cancer samples in the public domain as well as the post-diagnostic HGSOC samples collected in our preliminary cohort. Selected TCR features will be applied to develop a classifier for HGSOC patients by using the NHS sample cohort. The outcome of this Aim is a defined set of TCR sequences, or motifs, as a combined biomarker for early HGSOC.
Aim 2. TCR assay validation with PLCO prediagnostic samples. TCR-seq and HLA genotype data of all PLCO samples will be generated using the same platform or core facility to minimize technical artifacts. We will implement necessary preprocessing measures, such as sequencing depths, variable gene missing rate, TCR clonal richness etc, to select for high quality samples. Batch effects between cohorts will be evaluated using our TCR embedding method and controlled using a hierarchical empirical Bayes model. Multiple timepoints from a subset of individuals prior to diagnosis will be acquired to verify TCR repertoire dynamics. TCR biomarker in Aim 1 will be directly applicable to the data in this setting. To this end, we consider validation AUC≥0.85, with sensitivity greater than 37% at ≥99.5% specificity to be successful. This performance leads to a positive prediction value (PPV) of 5% for stage-I HGSOC, which is preliminarily useful for OC screening16.

Collaborators

Bo Li (Children's Hospital of Philadelphia)