Computer Identification of Unknown Spectra (Chemistry)

该项目是在数据开发和利用领域，涉及与化学文摘服务（CAS）合作，进一步提高计算机程序的准确性和速度，从它们的质量和红外光谱中识别未知化合物。这些程序是检索算法“基于概率的匹配（PBM）”和解释算法“自我训练解释和检索系统（STIRS）”。特别重要的是全面和有效地利用一个包含大约13万种不同质谱的新数据库。最高性能的PBM和STIRS程序将与CAS一起开发，利用他们在计算机结构操作方面的丰富经验和他们完整的连接表结构描述文件，最终用于他们的国际在线服务，STN。联机GC/MS的PBM将通过数据库压缩、预滤波技术、矢量公式和阵列处理器芯片的使用而变得更加高效。高分辨率GC/MS仪器提供的精确质量数据和红外光谱仪提供的结构信息将提高PBM和STIRS的精度。通过计算机识别最佳匹配化合物的最大结构描述，也将改进STIRS。

This project is in the area of data development and utilization and involves a collaborative effort with Chemical Abstracts Service (CAS) on further improvements in accuracy and speed in computer programs for identification of unknown compounds from their mass and infrared spectra. These programs are the retrieval algorithm "Probability Based Matching (PBM)" and the interpretive algorithm "Self-Training Interpretive and Retrieval System (STIRS)". Of particular importance will be comprehensive and efficient utilization of a new data base of around 130,000 different mass spectra. The highest performance PBM and STIRS programs will be developed with CAS, taking advantage of their extensive experience in computer structure manipulation and their complete files of connection-table structure descriptions, for ultimate use on their international on-line service, STN. PBM for on-line GC/MS will be made far more efficient with data base compression, pre-filter techniques, vector formulation, and use of array processor chips. PBM and STIRS accuracy will be improved by exact mass data available from high resolution GC/MS instruments and structural information from infrared spectrometers. STIRS will also be improved through computer identification of maximal structure descriptions of the best matching compounds.