Implementation of an expert system for the qualitative interpretation of X-ray fluorescence spectra

The implementation of an expert system for the automated qualitative interpretation of energy-dispersive x-ray spectra is discussed. The first step in the interpretation process is the extraction of the relevant data from the spectrum, which is done by a preprocessor program, written in FORTRAN. The expert system itself consists of three parts. The knowledge base contains specific information on energy-dispersive x-ray fluorescence spectrometry presented in the form of IF/THEN rules. The data base contains the reduced spectral data and an array of certainty factors associated with each element; the certainty factor for an element represents the probability of its being present in the sample from which the spectrum was taken. Finally, the inference engine performs manipulation of the knowledge. For a particular state of the data base, the certainty factors for all the elements are iteratively modified until convergence is reached by using the rules from the knowledge base. During each cycle, the inference engine selects one rule from the knowledge base and executes it. Rules are selected on the basis of the chemical elements contained in their IF part and according to their previously assigned focus levels. Execution of the THEN part of the selected rule modifies the certainty factors of a number of elements. At the end of the interpretation session, the system lists the elements which have a high probability of being present in the sample. Optionally, the user can be provided with explanations of the reasoning steps taken during the interpretation. Application of the expert system to a particular spectrum shows that it is useful for the reliable interpretation of spectral data obtained from electron microprobe analysis of industrial aerosol particles.     

Analysis of different synthetic homopolymers by the use of a new calculation software for tandem mass spectra

The manual interpretation of tandem mass spectra of synthetic polymers is very time-consuming. Therefore, a new software tool was developed to accelerate the interpretation of spectra obtained without requiring any further knowledge about the polymer class or the fragmentation behavior under high-energy collision-induced dissociation (CID) conditions. The software only requires an alphabetical list of elements and a peak list of the measured substance as an xml file for the evaluation of the chosen mass spectrum. Tandem mass spectra of different homopolymers, like poly(2-oxazoline)s, poly(ethylene glycol) and poly(styrene), were interpreted by the new software tool. This contribution describes a fast and automated software tool for the rapid analysis of homopolymers.     

Automated peak tracking for comprehensive impurity profiling in orthogonal liquid chromatographic separation using mass spectrometric detection

The presence and quantity of impurities in pharmaceutical drugs can have a significant impact on their quality and safety. With the continuous pressure for increased industry productivity, there is urgent need for a systematic and comprehensive drug impurity profiling strategy. We report here our development of the fully automated Comprehensive Orthogonal Method Evaluation Technology (COMET) system. The system includes five columns, seven orthogonal HPLC methods, and hyphenated UV-MS detections, which provides automated generic impurities screening for any drug sample. An automated MS peak tracking approach by program-based mass spectral interpretation is devised to unambiguously track impurities among all orthogonal HPLC methods. The program passes electro-spray ionization mass spectra (ESI-MS) through four sequential decision-making mass ion tests and determines molecular weights for every peak. The system reduces the time required to obtain impurity profile from weeks to days, while the automated MS peak tracking takes only minutes to interpret all MS spectral data of interest. Up-to-date, impurity contents of 56 in-development drug candidate samples have all been successfully illustrated by COMET, which contained more than 500 chemical entities. The program is able to track more than 80% of the compounds automatically with majority of the failure due to insufficient ionization for some impurities by ESI. This system is well suited for efficient drug development and ensuring the quality and safety of drug products.     

Expert system for interpretation of the infrared spectra of environmental mixtures

A program for the identification of the principal components of mixtures through interpretation of the infrared mixture spectrum (IntIRpret) was developed. This program, which was developed as a preliminary screening tool for unknown organic mixtures, has five main subroutines: the interferogram processing and peak-selection subroutine (PUSHSUB), the automated knowledge-acquisition subroutine (AUTOGEN), the system optimization subroutine (STO), the interpretation subroutine (PAIRS), and final processing subroutine to subtract spectral similarity (PAIRSPLUS). Principal advantages of this system compared to earlier systems are speed, flexibility and accuracy.     

Automated compound verification using 2D-NMR HSQC data in an open-access environment

Since the introduction of NMR prediction software, medicinal chemists have imagined submitting their compounds to corporate compound registration systems that would ultimately display a simplified pass/fail result. We initially implemented such a system based on HPLC and liquid chromatography mass spectrometry (LCMS) data that is embedded within our industry standard sample submission and registration process. By using gradient-heteronuclear single quantum coherence (HSQC) experiments, we have extended this concept to NMR data through a comparison of experimentally acquired data against predicted (1)H and (13)C NMR data. Integration of our compound registration system with our analytical instruments now provides our chemists unattended and automated NMR verification for collections of submitted compounds. The benefits achieved from automated processing and interpretation of results produced enhanced confidence in our compound library and released the chemists from the tedium of manipulating large amounts of data. This allows scientists to focus more of their attention to the drug discovery process.     

Peptide sequence determination from high-energy collision-induced dissociation spectra using artificial neural networks

This paper reports a newly developed technique that uses artificial neural networks to aid in the automated interpretation of peptide sequence from high-energy collision-induced dissociation (CID) tandem mass spectra of peptides. Two artificial neural networks classify fragment ions before the commencement of an iterative sequencing algorithm. The first neural network provides an estimation of whether fragment ions belong to 1 of 11 specific categories, whereas the second network attempts to determine to which category each ion belongs. Based upon numerical results from the two networks, the program generates an idealized spectrum that contains only a single ion type. From this simplified spectrum, the program’s sequencing module, which incorporates a small rule base of fragmentation knowledge, directly generates sequences in a stepwise fashion through a high-speed iterative process. The results with this prototype algorithm, in which the neural networks were trained on a set of reference spectra, suggest that this method is a viable approach to rapid computer interpretation of peptide CID spectra.     

Automated mass analysis of low-molecular-mass bacterial proteome by liquid chromatography-electrospray ionization mass spectrometry

Due to the presence of a large number of proteins in cell extracts, ion chromatograms of cell extracts obtained by electrospray ionization mass spectrometry (ESI-MS) can be quite complicated. It is found that the elevated baseline in an ion chromatogram contains many protein signals. One deficiency of current commercially available LC-ESI-MS data interpretation software is found to be the lack of functional operation that allows automated mass spectral integration and interpretation over signals hidden in the baseline. This current limitation can be overcome by a technique that involves the introduction of artificial pulses to an ion chromatogram by removing the solvent mixer in the HPLC pump. These artificial pulses are treated as chromatographic peaks by the software, thereby allowing automated spectral integration over the duration of a pulse. The reliability of mass analysis from the integrated spectra is shown to be dependent on spectral interpretation parameters such as mass spectral baseline threshold. The application of this method is demonstrated for rapid detection and mass analysis of low-molecular-mass proteins from cell extracts of Escherichia coli or Bacillus globigii.     

INFERCNMR: a 13C NMR interpretive library search system

INFERCNMR is an automated (13)C NMR spectrum interpretation aid for use either as a stand-alone program or as a component of a comprehensive, computer-based system for the characterization of chemical structure. The program is an interpretive library search which requires a database of assigned (13)C NMR spectra. An interpretive library search does not require overall structural similarity between an unknown and a library entry in order to retrieve a substructure common to both. Input consists of the chemical shift and one-bond proton-carbon multiplicity of each signal in the spectrum, and the molecular formula of the unknown. Program output is one or more substructures predicted to be present in the unknown, each of which is assigned an estimated prediction accuracy.     

Expertise — an expert system for infrared spectra evaluation

A system for the automated elucidation of chemical structures by the interpretation of infrared spectra is described. “Rules” for the finding of substructure features in an unknown spectrum are obtained by a pattern recognition procedure. A structure generation algorithm links particular substructures, called “superatoms”, together. Fuzzy set operators are used to consider inaccuracies in peak positions and intensities.     

Design and development of a new program for data processing of mass spectra acquired by means of a high-resolution double-focusing glow-discharge mass spectrometer

An new program has been developed and implemented for data analysis of mass spectra obtained by use of the VG9000 glow-discharge mass spectrometer. The program, designed to run in a Windows 9X environment includes several tools for import and export of data, cluster generators, etc. An automated technique for the interpretation of mass spectra is also built into the program; this enables faster and operator-independent interpretation. When major interferences or not-well defined signals are involved, the automated technique might fail to find the correct result. Therefore, a manual, VG9000 software-like, bypass is at hand. A comparison of the different techniques and programs shows, in general, comparable results. An installable version of the software is available on the university FTP-server (ftp://PLASMA-FTP.uia.ac.be/ private/imsas/).     

Automated data massaging, interpretation, and e-mailing modules for high throughput open access mass spectrometry

Hardware components and software modules were configured to enhance the automation, efficiency, and reliability of a commercial open access atmospheric pressure ionization mass spectrometry (API/MS) system for flow injection analysis. The data massaging module is a versatile package for data manipulation/reduction which is initialized upon detecting the end of data acquisition and can function in parallel during the data acquisition of the next sample. The data interpretation module compares the ions in the acquired mass spectrum with the predicted molecular adduct ions in different charge states, as well as the predicted isotopic distributions, possible artifact, polymer/cluster, byproduct/fragmentation ions, and then uses the results to score the quality of the spectrum. The e-mailing module transmits the spectrum and interpretation report to the desktop computer of the submitting chemist where the spectrum can be displayed and the report viewed. A scheme is also presented for the automated interpretation of an API mass spectrum for the determination of the most likely molecular weights of the components present in an “unknown” sample. Related flow diagrams, algorithms, and applications are illustrated.     

Artificial intelligence for the interpretation of combined spectral data : Design and development of a spectrum interpreter

The design of a knowledge-based system for the interpretation of combined spectral data for structure elucidation (EXSPEC) is described. Some basic design features are discussed and the functioning of the knowledge base, inference mechanism and user-interface is outlined. Attention is focussed on the development of a spectrum interpreter for infrared and mass spectral data. Interpretation of spectra for 120 liquid alcohols used for rule generation was successful. The system can be run on a Macintosh II or, more slowly, on a Macintosh Plus.     

Design and development of a new program for data processing of mass spectra acquired by means of a high-resolution double-focusing glow-discharge mass spectrometer

An new program has been developed and implemented for data analysis of mass spectra obtained by use of the VG9000 glow-discharge mass spectrometer. The program, designed to run in a Windows 9X environment includes several tools for import and export of data, cluster generators, etc. An automated technique for the interpretation of mass spectra is also built into the program; this enables faster and operator-independent interpretation. When major interferences or not-well defined signals are involved, the automated technique might fail to find the correct result. Therefore, a manual, VG9000 software-like, bypass is at hand. A comparison of the different techniques and programs shows, in general, comparable results. An installable version of the software is available on the university FTP-server (ftp://PLASMA-FTP.uia.ac.be/ private/imsas/).     

Divide-and-conquer,pattern matching,and relaxation methods in interpretation of 2-D NMR spectra of polypeptides

One task in the interpretation of the 2-D nuclear magnetic resonance (NMR) spectrum is to assign its signal patterns to their corresponding amino acids in proteins or polypeptides. To carry out this task of interpretation, one requires sufficient chemical knowledge and expertise to reason from a set of highly noisy data. We present a system called RUBIDIUM (a Rule-Based Identification in 2-D NMR Spectrum) to formulate the expertise and automate the process of interpretation. Given a protein or polypeptide with a known amino acid sequence and the 2-D NMR spectra (both COSY and NOESY), RUBIDIUM yields plausible assignments of lines that account for most signals observed in the spectrum and conform to prior chemical knowledge. Rules of pattern matching are used to detect plausible signal patterns. The expertise of the sequence-specific assignment task is formulated to assign a signal pattern to amino acids. To cope with ambiguities and noise, RUBIDIUM adopts various low-level data preprocessing techniques, the strategy of divide and conquer, and the relaxation technique to decrease the complexity and recover from overconstrained conditions. The polypeptides oxytocin and vasopressin are used to illustrate the performance of RUBIDIUM. © John Wiley & Sons, Inc.     

Computer-aided Interpretation of Proton Nuclear Magnetic Resonance Spectra (Ⅰ)——An Artificial Intelligence System for Interpretation of Proton NMR Spectra of Polymers

An artificial intelligence system for interpretation of proton NMR spectra of polymers is reported in this paper. The system, including spectra data base, knowledge data base and reasoning engine, is based on the characteristics of the proton NMR spectra of polymers and the spectra interpretation experiences of specialist. The system can partly simulate human thinking and interpret proton NMR spectra of polymers at different levels of sophistication. The program in the system was written in Turbo Prolog 2.0 and translated into machine language by computer compiler. It has been tested on an IBM PC/XT computer and a satisfactory result was given.     

Automated mass correction and data interpretation for protein open-access liquid chromatography-mass spectrometry

Characterization of recombinant protein purification fractions and final products by liquid chromatography-mass spectrometry (LC/MS) are requested more frequently each year. A protein open-access (OA) LC/MS system was developed in our laboratory to meet this demand. This paper compares the system that we originally implemented in our facilities in 2003 to the one now in use, and discusses, in more detail, recent enhancements that have improved its robustness, reliability, and data reporting capabilities. The system utilizes instruments equipped with reversed-phase chromatography and an orthogonal accelerated time-of-flight mass spectrometer fitted with an electrospray source. Sample analysis requests are accomplished using a simple form on a web-enabled laboratory information management system (LIMS). This distributed form is accessible from any intranet-connected company desktop computer. Automated data acquisition and processing are performed using a combination of in-house (OA-Self Service, OA-Monitor, and OA-Analysis Engine) and vendor-supplied programs (AutoLynx, and OpenLynx) located on acquisition computers and off-line processing workstations. Analysis results are then reported via the same web-based LIMS. Also presented are solutions to problems not addressed on commercially available, small-molecule OA-LC/MS systems. These include automated transforming of mass-to-charge (m/z) spectra to mass spectra and automated data interpretation that considers minor variants to the protein sequence-such as common post-translational modifications (PTMs). Currently, our protein OA-LC/MS platform runs on five LC/MS instruments located in three separate GlaxoSmithKline R&D sites in the US and UK. To date, more than 8000 protein OA-LC/MS samples have been analyzed. With these user friendly and highly automated OA systems in place, mass spectrometry plays a key role in assessing the quality of recombinant proteins, either produced at our facilities or bought from external sources, without dedicating extensive amounts of analyst resource.     

A totally automated data acquisition/reduction system for routine treatment of mass spectroscopic data by factor analysis

Target transformation factor analysis is applied to typical data from gas chromatography—mass spectrometry and solid-probe mass spectrometry to determine rapidly the number of components in unresolved or partially resolved peaks. This technique allows the detection of hidden impurities which often make interpretation or quantification impossible. The error theory of Malinowski is used to assess the reliability of the results. The totally automated system uses a commercially available g.c.—m.s. data system interfaced to a large computer, and the number of components under a peak can be determined routinely and rapidly.     

Simultaneous determination of selected endocrine disrupters (pesticides,phenols and phthalates) in water by in-field solid-phase extraction (SPE) using the prototype PROFEXS followed by on-line SPE (PROSPEKT) and analysis by liquid chromatography-atmospheric pressure chemical ionisation-mass spectrometry

In this study, a new procedure, based on on-line solid-phase extraction (SPE) and analysis by liquid-chromatography-atmospheric pressure chemical ionization-mass spectrometry (LC-APCI-MS), has been developed for the simultaneous, multianalyte determination of 21 selected pesticides, phenols and phthalates in water. SPE was carried out on polymeric PLRP-s cartridges by percolating 20 mL-samples. For sample preconcentration, the performance of a prototype programmable field extraction system (PROFEXS) was evaluated against the commercial laboratory bench Prospekt system used for method development. The Profexs is designed for the automated on-site sampling, SPE preconcentration, and storage of up to 16 samples in SPE cartridges. These cartridges are further eluted and on-line analyzed with the Prospekt coupled to the chromatographic system. In the optimized method, where completely on-line SPE-LC-MS analysis of the samples is carried out with the Prospekt in the laboratory, detection limits lower than 100 ng/L, and satisfactory precision (relative standard deviations <25%) and accuracies (recovery percentages >75%) were obtained for most investigated compounds from the analysis of spiked Milli-Q water. The extraction efficiency achieved with the Profexs was comparable to that of the Prospekt for most compounds and somewhat lower for the most apolar analytes, probably due to adsorption on the pump filters. The completely on-line optimized method was applied to the analysis of surface water, ground water and drinking water from a waterworks in Barcelona. Some pesticides and phenols were found in both surface water and groundwater at ng/L or µg/L levels, but not in the final drinking water. Di(2-ethylhexyl)phthalate (DEHP) was present in all samples investigated, including blanks. To the author's knowledge, this is the first work describing the application of a fully automated on-line SPE-LC-MS method for the simultaneous analysis of pesticides, phenols, and phthalates in water, and the second one that examines the possibilities of the prototype Profexs for automated on-site SPE preconcentration of organic pollutants from water samples.