Analysis of LC-MS data to identify peptide and glycan biomarkers for hepatocellul

描述(由申请人提供)：癌症的早期发现可以提高患者的存活率。表征多肽和多糖与癌症的关联是发现早期诊断癌症生物标志物的最有前途的策略之一。本研究利用液质联用(LC-MS)技术对慢性肝病(CLD)向肝细胞癌(肝细胞癌)发展过程中的多肽和多糖表达谱进行了研究。我们的目标是在CLD高危人群中寻找和验证在可治疗阶段检测肝癌的多肽和葡聚糖生物标志物。无标记LC-MS定量允许以良好的吞吐量比较多肽和多糖，这使得我们能够比较大量的患者。然而，这种量化在特定于仪器的软件包中没有得到充分的处理。特别是，LC-MS数据的比对和标准化对生物分子的无标记定量和比较提出了重大挑战。这一挑战加上人类群体中的生物变异性和疾病异质性，限制了基于LC-MS的生物标记物发现研究的最新进展。该项目汇集了生物信息学、生物统计学、生物化学和质谱学的专家，开发了一套基于LC-MS的新型分析工具，用于血清和血浆中多肽和多糖的无标记定量和比较。具体地说，将研究一种新的贝叶斯分层模型，用于LC-MS数据的同时对齐和归一化，以及用于识别患者亚组。贝叶斯框架涉及固定和随机效应，以解释亚群的同质行为(固定的系统变化)，同时允许对群体内的异质性进行建模(随机效应)。将进行一项尖峰研究，以获得已知多肽和多糖浓度的重复LC-MS运行。这些数据将被用来开发和优化拟议的贝叶斯框架，并将其性能与其他现有解决方案进行比较。优化的框架和基于机器学习的特征选择方法将被应用于识别一套整合的多肽和多糖候选生物标记物，用于肝癌的早期检测。LC-MS分析肝癌患者血清和血浆中的整合肽和多糖是我们所知的史无前例的。将使用埃及和美国肝细胞癌和慢性阻塞性肺疾病患者的血液样本。这些生物标志物将使用同位素稀释质谱分析进行验证。

DESCRIPTION (provided by applicant): Early detection of cancer improves patient survival. Characterizing the association of peptides and glycans with cancer is one of the most promising strategies to discover early-diagnosis cancer biomarkers. This study evaluates peptide and glycan expression profiles in the progression of chronic liver disease (CLD) to hepatocellular carcinoma (HCC) by using the liquid chromatography-mass spectrometry (LC-MS) technology. The goal is to find and validate peptide and glycan biomarkers for detection of HCC at a treatable stage in a high-risk population of patients with CLD. Label-free LC-MS quantification allows comparison of peptides and glycans with good throughput which allows us to compare a large population of patients. However, such quantification is not addressed adequately in the instrument-specific software packages. In particular, alignment and normalization of LC-MS data present a significant challenge in label-free quantification and comparison of biomolecules. This challenge coupled with biological variability and disease heterogeneity in human populations has restricted recent advances in LC-MS-based biomarker discovery studies. This project brings together experts in bioinformatics, biostatistics, biochemistry, and mass spectrometry to develop a suite of novel analytical tools for LC-MS-based label-free quantification and comparison of peptides and glycans in serum and plasma. Specifically, a novel Bayesian hierarchical model will be investigated for simultaneous alignment and normalization of LC-MS data and for identification of patient subgroups. The Bayesian framework involves fixed and random effects to account for subpopulation homogeneous behavior (fixed systematic changes), while allowing for modeling heterogeneity within a group (random effects). A spike-in study will be conducted to obtain replicate LC-MS runs with known peptide and glycan concentrations. The data will be utilized to develop and optimize the proposed Bayesian framework and to compare its performance with other existing solutions. The optimized framework and a machine learning-based feature selection method will be applied to identify an integrated set of peptide and glycan candidate biomarkers for early detection of HCC. LC-MS analysis of integrated peptides and glycans in both serum and plasma of patients with HCC is to our knowledge unprecedented. Blood samples from patients with HCC and CLD controls in Egypt and United States will be used. The biomarkers will be validated using isotope dilution mass spectrometric assays.