COMPUTATIONAL TOOLS FOR CANCER PROTEOMICS

描述（由申请人提供）： 该申请的目标是开发新的计算方法来分析蛋白质表达和磷酸化变化对信号通路和疾病状态的反应，直接支持三个合作者实验室中进行的黑色素瘤和前列腺癌研究。使用基于肽气相片段化的多维LC/MSMS方法的鸟枪蛋白质组学，如MuDPIT，已被证明在复杂样品中的蛋白质的分离中是有效的。然而，在复杂混合物中的蛋白质取样深度、将肽序列分配至MSMS谱的准确性、区分蛋白质异构体的模糊性、蛋白质丰度的定量和翻译后修饰（如磷酸化）的表征方面存在严重的限制。此外，需要方法来处理复杂混合物产生的问题，例如质量和洗脱重叠的峰，多维分离期间在许多级分中洗脱的肽，以及基于多变量测量的肽/蛋白质的聚类。拟议的实验将开发新的计算工具，以创建一个集成的软件系统，将解决这些目标。具体目标是：（1）开发计算工具，用于量化多维LC分级样品中蛋白质丰度的变化，（2）通过改进理论MS/MS光谱预测算法来提高肽和蛋白质鉴定的准确性，（3）开发统计和计算方法，以改进复杂样品中的磷酸肽分析，以及（4）开发图像识别神经网络策略，用于在多个样本之间的多维数据集中聚类肽和磷酸肽特征。这些目标的完成将解决鸟枪法蛋白质组学中悬而未决的障碍，并提供强大的计算工具，以实现准确和灵敏的蛋白质分析，差异磷酸化的评估，多个平台和样本的多变量数据集的整合，以及快速描绘蛋白质组学数据集中疾病判别因子的新算法。随着这些工具的开发，它们将被应用于三个项目，涉及基础和临床癌症研究的蛋白质组学，分析癌细胞，组织和液体中的分子变化，以发现癌症生物标志物。所有三个项目的数据收集将使用我们生物分子质谱核心设施中的LTQ-Orbitrap和4000 QTrap质谱仪器进行，研究人员将访问正在开发的软件进行数据简化。这将提供来自调查人员的关于结果和经验的持续反馈，这将使团队能够通过故障排除软件和增加进一步的分析能力来应对现实世界样品的需求。

DESCRIPTION (provided by applicant): The goal of this application is to develop new computational methods to profile protein expression and phosphorylation changes in response to signaling pathways and disease states, directly supporting studies of melanoma and prostate cancer carried out in the laboratories of three collaborators. Shotgun proteomics using multidimensional LC/MSMS approaches that are based on peptide gas phase fragmentation, such as MuDPIT, have proven effective in idenfitying proteins in complex samples. However, there are serious limitations with respect to depth of sampling proteins in complex mixtures, accuracy of assigning peptide sequences to MSMS spectra, ambiguities in distinguishing protein isoforms, quantification of protein abundances, and characterization of posttranslational modifications, such as phosphorylation. In addition, methods are needed to handle problems arising with complex mixtures, such as peaks that overlap in mass and elution, peptides eluting in many fractions during multidimensional separation, and clustering of peptides/proteins based on multivariate measurements. The proposed experiments will develop new computational tools to create an integrated software system which will address these goals. The specific aims are to (1) develop computational tools for quantifying changes in protein abundances from samples fractionated by multidimensional LC, (2) increase the accuracy of peptide and protein identifications by improving algorithms for theoretical MS/MS spectral predictions, (3) develop statistical and computational methods to improve phosphopeptide analyses in complex samples, and (4) develop an Image Recognition Neural Network strategy for clustering peptide and phosphopeptide features within multidimensional datasets between many samples. Completion of these aims will address outstanding unsolved obstacles in shotgun proteomics and provide robust computational tools to achieve accurate and sensitive protein profiling, assessment of differential phosphorylation, integration of multivariate datasets from multiple platforms and samples, and new algorithms for rapid delineation of disease discriminators in proteomics datasets. As these tools are developed, they will be applied to three projects involving proteomics for basic and clinical cancer research, profiling molecular changes in cancer cells, tissues, and fluids for cancer biomarker discovery. Data collection for all three of projects will be carried out using LTQ-Orbitrap and 4000 QTrap mass spectrometry instruments available in our biomolecular mass spectrometry core facility, where investigators will access the software under development for data reduction. This will provide continual feedback from investigators about results and experiences, which will allow the team to respond by troubleshooting software and adding further analytical capabilities for the needs of real-world samples.