Intelligent Aids for Proteomic Data Mining

描述(由申请人提供)：本建议的主要目的是为申请人提供实现以下两个目标的手段和结构：(1)开发智能计算辅助工具，用于挖掘从高通量技术(如SELDI-TOF质谱仪)中积累的蛋白质组数据；以及(2)长期目标是通过发展贝叶斯方法和蛋白质组技术的方法学专业知识，获得作为生物医学信息学研究人员的独立地位。申请者将获得数据分析的概率方法的进一步指导；她将接受推动当今蛋白质组研究的蛋白质组技术的教育。除了由优秀导师指导的研究外，还将通过正式的课程作业、定向阅读、研讨会和会议提供培训。 申请者的研究项目涉及一种用于蛋白质组数据分析的新技术组合。以前的研究包括使用遗传算法和神经网络等技术来分析蛋白质组数据。这些技术的设计没有明确考虑到背景和先验知识。该项目的假设是，背景知识和机器学习技术可以积极影响从蛋白质组数据中选择合适的生物标记物，从而能够高效和准确地分析来自蛋白质组图谱研究的海量数据集。因此，这个项目将满足四个目标：(1)开发基于WRAPPER的机器学习工具；(2)利用先验知识，如数据中的启发式规则和关系来增强该工具；(3)将这些特征与未识别的患者信息一起用作分类系统的输入；以及(4)评估用于解释串联质谱学(MS-MS或MS/MS)数据的现有技术，并提出、实施和评估用于识别MS-MS谱指示的多肽和蛋白质的贝叶斯方法。

DESCRIPTION (provided by applicant): Primary purpose of this proposal is to provide the applicant with the means and structures for achieving two goals; (1) to develop intelligent computational aids for mining proteomic data accumulating from high throughput techniques like SELDI-TOF mass spectrometry; and (2) the long-term goal is to gain independence as a biomedical informatics researcher by developing methodological expertise in Bayesian methods and proteomic technologies. Applicant will obtain further instruction in probabilistic methods of data analysis; and she will receive education on proteomic technologies that are driving today's proteome research. Training will be provided through formal coursework, directed readings, seminars and conferences in addition to research directed by excellent mentors. Applicant's research project involves a novel combination of techniques for use in proteomic data analysis. Previous research has included the use of techniques such as genetic algorithms and neural networks for analysis of proteomic data. These techniques were not explicitly designed to take into account background and prior knowledge. Hypothesis of this project is that background knowledge and machine learning techniques can positively influence the selection of appropriate biomarkers from proteomic data, enabling efficient and accurate analysis of massive datasets arising from proteomic profiling studies. Therefore, this project will satisfy four aims: (1) development of a wrapper-based machine learning tool; (2) augment the tool with prior knowledge such as heuristic rules and relationships in the data; (3) use these features along with de-identified patient information as input to classification systems; and (4) evaluate existing techniques for interpreting tandem mass spectrometry (MS-MS or MS/MS) data, and propose, implement and evaluate a Bayesian method for identification of peptides and proteins indicated by the MS-MS spectrum.