CCRC NET 1H NMR SPECTRAL DATABASE OF N LINKED OLIGOSACCHARIDES

N 连接低聚糖的 CCRC NET 1H NMR 谱数据库

• 批准号: 6653526
• 负责人: FARAMARZ VALAFAR
• 金额: $ 28.8万
• 依托单位: UNIVERSITY OF GEORGIA
• 依托单位国家: 美国
• 财政年份: 2002
• 资助国家: 美国
• 起止时间: 2002-08-01 至 2003-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6653526
• 关键词: biological products; biomedical resource; carbohydrates; computers; informatics; model design /development; nuclear magnetic resonance spectroscopy; technology /technique

A sixth database was added to the CCRC-Net this year to increase the resources for carbohydrate structural analysis and identification provided by the CCRC Web site. This is the second database developed for N-linked oligosaccharides. The neural network-based search engine created for this database was successfully trained and developed using real spectra and is a far more robust search engine than we had been using previously. Our experiments to establish a methodology for a more robust search engine indicated that multiple spectra for each structure are needed during training of the artificial neural networks (ANNs). We used at least three spectra for each of 23 structures for training the ANNs. We found that this engine reports a false-positive much less frequently than our standard engine for N-linked oligosaccharides. The new engine can also tolerate an additional 10% decrease in the signal-to-noise ratio.

今年，CCRC-Net增加了第六个数据库， 碳水化合物结构分析和鉴定资源 由CCRC网站提供。 这是开发的第二个数据库 对于N-连接的寡糖。 基于神经网络的搜索引擎 为该数据库创建的数据库已成功地使用 真实的光谱，是一个比我们以前强大得多的搜索引擎 使用以前。 我们的实验旨在建立一种方法论 更强大的搜索引擎表明，多个光谱，为每一个 在人工神经网络的训练过程中， （ANN）。 我们对23个结构中的每一个使用了至少三个光谱， 训练ANN。 我们发现这个引擎报告了一个假阳性 比我们的标准引擎的N-连接频率低得多 低聚糖。 新发动机还可以承受额外的10% 降低信噪比。