Developing a Platform for Prediction of Metastasis Using Multiplexed QD-imaging

描述（由申请人提供）：多元生物标志物分析比单一生物标志物分析更能反映肿瘤的生物学行为，但其标准化和量化仍然是一个挑战。此外，大多数计算机软件不提供成像和分析生物标志物亚细胞定位的方法，并将它们与生物学和临床信息联系起来。本项目的目的是开发一个结合多路免疫染色成像和定量以及生物信息学的平台，用于预测头颈部鳞状细胞癌（SCCHN）原发肿瘤（PT）的淋巴结转移（LNM）。SCCHN的LNM是一种精确定义的生物现象，是开发多路生物标志物平台（MBP）的理想模型。基于我们的初步研究，我们的目标是验证假设，即MBP可以用于同时识别多种转移相关生物标志物在PTs中的亚细胞分布和表达。这些生物标志物的准确定量将有助于预测PTs的转移。三种新兴技术，基于量子点（QD）的免疫组织荧光（IHF），多光谱成像和机器学习将用于验证这一假设。利用这些方法，将开发一个将定量多重免疫染色与生物统计学相结合的平台，并对其敏感性、特异性和临床预测能力进行测试。因此，本项目非常适合NCI项目公告“癌症预后和预测的发展研究”（PA-09-159）的范围。本研究提出了三个目标。(1)建立基于大体积组织模型的多重生物标志物系统和方法，用于预测SCCHN PT组织中LNM的发生。本研究将建立并验证一种分析方法，利用InForm软件中的新功能对膜和细胞质染色进行多重定量分析，其中将具体分析某些生物标志物的亚细胞定位。基于该体组织模型的LNM预测将实现。(2)建立基于亚群体模型的单细胞定量方法，用于预测SCCHN PT组织中LNM的发生。每个细胞的分析结果将被量化为相同PTs中多重生物标志物分析中高风险细胞的百分比。高危人群与LNM相关。将亚种群模型预测的敏感性和特异性与体组织模型进行比较。(3)结合转移的临床特征，利用软件开发并验证一种nomogram，作为预测LNM的工作平台。虽然Aim 1和Aim 2的主要目的是将这三种生物标志物与转移联系起来，但其他临床因素如分化状态、肿瘤分期、部位等也可能与LNM相关。最具预测性的生物标志物集与相关临床因素将构成一个计算机软件平台，将在另外100个SCCHN样本中进行验证，以预测LNM。

DESCRIPTION (provided by applicant): Multiplexed biomarker analysis is more powerful in reflecting the biological behaviors of a tumor than single biomarker analysis, but its standardization and quantification is still a challenge. Furthermore, most computer software does not provide methods for imaging and analyzing subcellular localization of biomarkers and correlating them with biological and clinical information. The objective of this project is to develop a platform which combines imaging and quantification of multiplexed immunostaining plus bioinformatics for the prediction of lymph node metastases (LNM) from the primary tumor (PT) of squamous cell carcinoma of the head and neck (SCCHN). LNM of SCCHN is a precisely defined biological phenomenon which is an ideal model to be utilized to develop this multiplexed biomarker platform (MBP). Based on our preliminary studies, we aim to test the hypothesis that that the MBP can be developed to identify the subcellular distribution and expression of multiple metastasis-related biomarkers simultaneously in PTs. Accurate quantification of these biomarkers will facilitate the prediction of metastasis from PTs. Three emerging technologies, quantum dot (QD)-based immunohistofluorescence (IHF), multispectral imaging, and machine learning will be used to test this hypothesis. Using these approaches, a platform that combines quantifying multiplexed immunostaining with biostatistics will be developed and tested for its sensitivity, specificity, and prediction power for use in the clinic. Therefore, this project fits appropriately to the scope of the NCI program announcement "Developmental Research in Cancer Prognosis and Prediction" (PA-09-159). Three aims are proposed in the study. (1) To develop a multiplexed biomarker system and method based on a bulk tissue model for prediction of LNM in SCCHN PT tissues. This Aim will establish and validate an analysis methodology for multiplexed quantification of membrane and cytoplasmic staining using a new function in InForm software where subcellular localization of certain biomarkers will be specifically analyzed. Prediction of LNM based on this bulk tissue model will be achieved. (2) To develop a per-cell quantification method based on a sub-population model for prediction of LNM in SCCHN PT tissue. The per-cell analysis results will quantified as the percentage of high risk cells from the multiplexed biomarker analyses in the same PTs. The high risk population will be correlated with LNM. The sensitivity and specificity of the prediction by the sub-population model will be compared with that of the bulk tissue model. (3) To develop and validate a nomogram with software combining clinical characterizations of metastasis as a working platform for the prediction of LNM. While the primary endpoint of Aim 1 and 2 is to correlate the three biomarkers with metastasis, other clinical factors such as differentiation status, tumor stage, and site, etc. may also correlate with LNM. The most predictive biomarker set combined with relevant clinical factors will constitute a platform with computer software that will be validated in an additional 100 SCCHN samples for prediction of LNM.