Joint Neural Network Interpretation of Infrared and Mass Spectra

Abstract

Combining gas phase infrared (IR) spectra with mass spectral (MS) data, a neural network has been developed to predict 26 different molecular substructures from multispectral information. The back-propagation procedure has been used for training, including its previously published modification, the flashcard algorithm. Present functional groups have been detected correctly in 86.4% of all cases, compared with 88.4% using only IR and 78.2% using only MS data for training and prediction. For only 8 out of the 26 functionalities does the joint utilization of infrared and mass spectra yield better prediction results, with the greatest improvement being for halogen bond predictions. The prediction of functional group absence results in accuracy of about 95.5% for both IR and IR/MS networks but only 87.1% for a stand alone MS network. Insights have been gained into the suitability of both data sets for neural network training by presenting just IR or MS data to a jointly trained neural network, revealing the amount of information the network utilizes from either spectroscopic technique. In addition, an algorithm which produces balanced training and test sets for multi-output neural networks has been devised.

Exploration of anthropological specimens by GC‐MS and chemometrics

“…Several of the very first papers that tried to use multivariate data analysis for chemical problems deal with the automatic recognition of substance classes from low resolution mass spectra (Crawford & Morrison, 1968;Jurs, Kowalski, & Isenhour, 1969). Classification of the presence or absence of substructures based on mass spectral data has found great interest and has been suggested as a tool in systematic structure elucidation of organic molecules (Chapman, 1993;Klawun & Wilkins, 1996;Varmuza & Werther, 1996;Varmuza, 2000). Recently, exploratory data analysis methods have proven to be useful in the evaluation of database search results or spectra similarity hitlists, as well as for the recognition of spectra-structure relationships Scsibrany et al, 2003).…”

Maximum Common Substructures of Organic Compounds Exhibiting Similar Infrared Spectra

“…Besides NMR spectroscopic techniques and mass spectrometry the use of infrared (IR) spectral data plays an important role in structure elucidation. 1,2 Computer-based approaches for the interpretation of IR spectra can be classified into three categories: (1) knowledge-based systems in which chemical expertise is encoded to assist in spectra interpretation [3][4][5] ; (2) pattern recognition methods based on multivariate data analysis, statistics, and neural networks [6][7][8][9] ; and (3) the most widely used technique, namely search in spectral libraries. Each of these approaches has its own advantages and limitations, but especially library search methods have demonstrated their usefulness in scientific and laboratory practice.…”

The American Society for Mass Spectrometry

“…This paper focuses on the application of multivariate data analysis -the typical chemometric approach -to investigate relationships between low resolution electron impact data and chemical structures as well as on the development and use of MS classifiers together with automatic isomer generation [6]. Chemometric methods are successful to some extent in the automatic recognition of substructures or other structural properties from low resolution electron impact mass spectra [6][7][8][9][10][11][12][13][14][15]. In some cases a systematic structure elucidation is possible from the molecular formula of an unknown together with restrictions about the presence or absence of substructures (automatically obtained from spectra).…”