Thomas P. Forrest

Neural network approach to structural feature recognition from infrared spectra

Neural networks, with and without hidden nodes, have been trained to recognize structural features of compounds from their infrared spectra. The training of the networks was evaluated by a variety of statistical indices using threshold values obtained by simplex optimization and by evaluation of synthetic spectra of structural groups obtained from the connection weights of the single-layer networks. Results indicate that all of the networks can be trained to recognize the structural groups in the compounds used to train the network. The network with a hidden layer, and dedicated to a single structural group, was better able to recognize structural groups in compounds that had not been used in training the network. Although not as efficient, the single-layer networks are particularly useful in that information may be extracted for use in writing more effective rules for an expert system-based infrared interpreter.

Identification of structural features from mass spectrometry using a neural network approach: application to trimethylsilyl derivatives used for medical diagnosis.

An artificial neural network (ANN) has been trained to recognize the presence or absence of specific structural features (SF) in trimethylsilyl derivatives of organic acids from their mass spectra. The input vector is constructed without knowledge of the molecular ion, which is generally not observed in the spectra of these compounds. The results are used in conjunction with a classical search in a spectral library to identify organic acids in biological fluids for rapid acidemias diagnosis.

Metabolism of primary bile acids by Clostridium perfringens

Abstract C. perfringens degraded chenodeoxycholic acid (CDC) into two major products: 7α-hydroxy-3-oxo-5β-cholanoic acid and 3β,7α-dihydroxy-5β-cholanoic acid, the latter appearing very quickly and the former very slowly in whole cell cultures. Yields of about 60 and 16% were obtained for the respective products. Both products, as substrates in C. perfringens cultures, were quickly transformed back to CDC (about 84%) and 3β,7α-dihydroxy-5β-cholanoic acid (about 16%), whereas 7α-hydroxy-3-oxo-5β-cholanoic acid was formed slowly (4–5 days). In contrast to CDC, cholic acid was not epimerized at 3α-OH position; however, 7α, 12α-dihydroxy-3-oxo-cholanoic acid was formed slowly. When 7α,12α-dihydroxy-3-oxo cholanoic acid was substrate, rapid quantitative reduction, back to cholic acid then slow oxidation back to the 3-oxo product was observed. Thirty-six strains of C. perfringens (18 from a culture collection and 18 isolated from human feces) were screened for 3α-OH epimerization and oxidation of the primary bile acids. All but three strains epimerized CDC (yields 14–20%), while none epimerized cholic acid. Oxidation at the 3 position varied from 0–30% for CDC and 0–90% for cholic acid. Cell-free preparations of C. perfringens in the presence of NADP produced both 3-oxo and 3β-hydroxyl products with CDC, but only the 3-oxo product with cholic acid; these activities were non-inducible. The authors propose that 3α-OH isomerization takes place early in log phase via a 3α- and 3β-hydroxysteroid dehydrogenase (HSDH) system. Accumulation of 3-oxo product occurs late in stationary phase as the Redox (Eh) value rises above —100 mV.

Development of neural networks for identification of structural features from mass spectral data.

In this paper we present a methodology for developing optimized artificial neural network (ANN) and describe the application of this methodology to the creation of a network specialized in the identification of the structural features of compounds from mass spectral data. We also make a critical comparison of the results obtained from the application of our methodology with those obtained by other authors who have explored similar problems with or without using predefined methodologies. We believe that our methodology and the tools needed to refine the applications can be used in other chemical problems involving the identification of structural features of compounds.

The Enzymic and Chemical Synthesis of Ursodeoxycholic and Chenodeoxycholic Acid from Cholic Acid

Three approaches to the synthesis of ursodeoxycholic acid (UDC) from cholic acid have been investigated: (i) oxidation of cholic acid to 3 alpha, 7 alpha-dihydroxy-12 keto-5 beta-cholanoic acid (12K-CDC) with Clostridium group P 12 alpha-hydroxysteroid dehydrogenase (HSDH), isomerization of 12K-CDC to 3 alpha, 7 beta-dihydroxy-12 keto-5 beta-cholanoic acid (12K-UDC) with Clostridium absonum 7 alpha- and 7 beta-HSDH and reduction of 12K-UDC by Wolff-Kishner to UDC; (ii) isomerization of cholic acid to ursocholic acid (UC) by C. absonum 7 alpha- and 7 beta-HSDH, oxidation of UC to 12K-UDC with Clostridium group P 12 alpha-HSDH and Wolff-Kishner reduction of 12K-UDC to UDC; (iii) oxidation of cholic acid to 12K-CDC by Clostridium group P 12 alpha-HSDH, Wolff-Kishner reduction of 12K-CDC to chenodeoxycholic acid (CDC) and isomerization of CDC to UDC using whole cell cultures of C. absonum. In the first two approaches (using cell free systems) the yields of desired product were relatively low primarily due to the formation of various side products. The third method proved the most successful giving an overall yield of 37% (UDC) whose structure was verified by mass spectroscopy of the methyl ester.

Advantages of a hierarchical system of neural-networks for the interpretation of infrared spectra in structure determination

Abstract A hierarchical system of small feed forward neural-networks is used to extract structural information from infrared spectra. The top-level network gives a rough classification in five non-exclusive classes: compounds containing carbonyl, hydroxyl, amino groups, aromatic compounds and ethylenic compounds. For each class, a dedicated network is designed to identify more specific structural features. Depending upon those structural features, the hierarchy might be extended to deeper levels. Specialised networks are activated in a cascade-like effect by the output of the upper-level networks. The training of each specialist network is performed using learning and test sets made of compounds identified by the upper level networks as belonging to this class. Thanks to this approach and to the optimisation of decision thresholds, the quality of the responses is excellent, and compounds wrongly classified by one network do not lead automatically to other errors. One major advantage of this approach is the limited size of each network involved. Networks with few outputs are easier to optimise, and their performance is better than that of larger networks. Moreover linking the response sets from the different refinement levels allows improvement of response quality and in some cases inference of other structural features by combination of responses. Hierarchical neural-network systems are well suited for the interpretation of infrared spectra. They perform very well, and the different refinement levels of information permit great flexibility in the ways they may be used. The modular organisation allows modification of some parts of the system without damaging the whole hierarchy.

Indices for the evaluation of neural network performance as classifier: Application to structural elucidation in infrared spectroscopy

An application of Artificial Neural Networks (ANN) to the substructure detection, from infrared spectra, of organic compounds is described. Several ANNs have already been implemented for this purpose, and show promising initial results; however, many problems remain to be resolved. We wished to train ANNs to assist in a decision support system using several spectroscopic methods to elucidate the structure of unknown molecules. To optimise the ANN with respect to spectral feature extraction, network architecture, training regime and threshold determination, we have investigated several indices for use in the evaluation of network performance. Since much published work on ANN application in this field present performance indices that are poorly defined or of limited use, we recommend that the basic results be reported so that readers may calculate indices to suit their own particular needs. These basic quantities are identified and a set of derived indices recommended.

Identification of 7α-,12α-dihydroxy-3-oxo cholanoic acid as the major degradation product from cholic by C. perfringens

Abstract The major degradation product of cholic acid by C. perfringens, was identified as 3-dehydrocholic acid by n.m.r. and mass spectroscopic analysis of its methyl ester. 3-Dehydrocholic acid was shown to be unreactive with P. testosteroni 3α-hydroxysteroid dehydrogenase (forward direction) and relatively unreactive with 7α-hydroxysteroid dehydrogenase. Thus, both 3α-OH and 7α-OH bioconversion determinations in spent bacterial media have been shown to be useful in estimating the amount of 3-dehydrocholic acid formed from cholic acid in whole cell cultures of 63 strains of C. perfringens.

Clean-up of Spectro-Structural Databases using Artificial Neural Networks

Quality control in the construction and usage of factual databases is a well known problem. Classical methods of data inspection are not adequate for spectro-structural data of a large number of molecular spectra of various origins (IR, MS, NMR, etc.). Artificial neural networks can be successfully used as non-linectr mapping devices between spectroscopic and structural features. We have built and trained hierarchical neural networks to recognize the presence or absence of several functional groups in a molecule from its infrared spectra Thanks to the speed of computation of these neural networks, it is possible to scan a large spectro-structural database in order to identify doubtful spectra and/or structures. Typically, these represent only two to 15% of the records.