Pattern recognition analysis of 1H NMR spectra from perchloric acid extracts of human brain tumor biopsies

Pattern recognition techniques (factor analysis and neural networks) were used to investigate and classify human brain tumors based on the 1H NMR spectra of chemically extracted biopsies (n = 118). After removing information from lactate (because of variable ischemia times), unsupervised learning suggested that the spectra separated naturally into two groups: meningiomas and other tumors. Principal component analysis reduced the dimensionality of the data. A back-propagation neural network using the first 30 principal components gave 85% correct classification of meningiomas and nonmeningiomas. Simplification by vector rotation gave vectors that could be assigned to various metabolites, making it possible to use or to reject their information for neural network classification. Using scores calculated from the four rotated vectors due to creatine and glutamine gave the best classification into meningiomas and nonmeningiomas (89% correct). Classification of gliomas (n = 47) gave 62% correct within one grade. Only inositol showed a significant correlation with glioma grade.     

Assessment of statistical and neural networks methods in NMR spectral classification and metabolite selection

Magnetic resonance spectroscopy opens a window into the biochemistry of living tissue. However, spectra acquired from different tissue types in vivo or in vitro and from body fluids contain a large number of peaks from a range of metabolites, whose relative intensities vary substantially and in complicated ways even between successive samples from the same category. The realization of the full clinical potential of NMR spectroscopy relies, in part, on our ability to interpret and quantify the role of individual metabolites in characterizing specific tissue and tissue conditions. This paper addresses the problem of tissue classification by analysing NMR spectra using statistical and neural network methods. It assesses the performance of classification models from a range of statistical methods and compares them with the performance of artificial neural network models. The paper also assesses the consistency of the models in selecting, directly from the spectra, the subsets of metabolites most relevant for differentiating between tissue types. The analysis techniques are examined using in vitro spectra from eight classes of normal tissue and tumours obtained from rats. We show that, for the given data set, the performance of linear and non-linear methods is comparable, possibly due to the small sample size per class. We also show that using a subset of metabolites selected by linear discriminant analysis for further analysis by neural networks improves the classification accuracy, and reduces the number of metabolites necessary for correct classification.     

High sodium intake decreases pressure-induced (myogenic) tone and flow-induced dilation in resistance arteries from hypertensive rats

A real-time automated way of quantifying metabolites from in vivo NMR spectra using an artificial neural network (ANN) analysis is presented. The spectral training and test sets for ANN containing peaks at the chemical shift ranges resembling long echo time proton NMR spectra from human brain were simulated. The performance of the ANN constructed was compared with an established lineshape fitting (LF) analysis using both simulated and experimental spectral data as inputs. The correspondence between the ANN and LF analyses showed correlation coefficients of order of 0.915-0.997 for spectra with large variations in both signal-to-noise and peak areas. Water suppressed 1H NMR spectra from 24 healthy subjects were collected and choline-containing compounds (Cho), total creatine (Cr), and N-acetyl aspartate (NAA) were quantified with both methods. The ANN quantified these spectra with an accuracy similar to LF analysis (correlation coefficients of 0.915-0.951). These results show that LF and ANN are equally good quantifiers; however, the ANN analyses are more easily automated than LF analyses.     

Quantification of biomedical NMR data using artificial neural network analysis: lipoprotein lipid profiles from 1H NMR data of human plasma

Artificial neural network (ANN) analysis is a new technique in NMR spectroscopy. It is very often considered only as an efficient "black-box' tool for data classification, but we emphasize here that ANN analysis is also powerful for data quantification. The possibility of finding out the biochemical rationale controlling the ANN outputs is presented and discussed. Furthermore, the characteristics of ANN analysis, as applied to plasma lipoprotein lipid quantification, are compared to those of sophisticated lineshape fitting (LF) analysis. The performance of LF in this particular application is shown to be less satisfactory when compared to neural networks. The lipoprotein lipid quantification represents a regular clinical need and serves as a good example of an NMR spectroscopic case of extreme signal overlap. The ANN analysis enables quantification of lipids in very low, intermediate, low and high density lipoprotein (VLDL, IDL, LDL and HDL, respectively) fractions directly from a 1H NMR spectrum of a plasma sample in < 1 h. The ANN extension presented is believed to increase the value of the 1H NMR based lipoprotein quantification to the point that it could be the method of choice in some advanced research settings. Furthermore, the excellent quantification performance of the ANN analysis, demonstrated in this study, serves as an indication of the broad potential of neural networks in biomedical NMR.     

Patterns of occurrence of second primary non-mammary malignancies in breast cancer patients: results from 1382 consecutive autopsies

In a previous study, we demonstrated the existence of a 3.2 +/- 0.2 ppm peak in the 1H NMR spectrum at 60 MHz from human pancreatic adenocarcinomas (Capan-1 cell) heterotransplanted into nude mice. This peak, which is not present in normal human pancreas, was attributed to enhanced membrane fluidity and/or or an increase in phospholipid turnover. The present study was designed to identify this signal by comparing the 1H NMR spectra recorded in vivo at 100 MHz from Capan-1 tumors, after suppression of the tissular water proton peak, to those recorded from normal pancreatic tissue, and to those recorded at 300 MHz from lipid extracts. The 1H NMR spectra at 100 MHz of the Capan-1 tumors in vivo exhibited three main peaks in the 3.2 +/- 0.2 ppm region: 1. A peak at 2.8 +/- 0.1 ppm from CH2 protons of the acyl chains of unsaturated phospholipids; 2. A peak at 3.2 +/- 0.1 ppm from the protons of the N(CH3)3 group of choline; and 3 A peak at 3.5 +/- 0.1 ppm attributed to GPC. The NMR 1H 300 MHz spectrum of phospholipid extracts of Capan-1 tumors displayed 12 principal resonances, of which only the N(CH3)3 peak of PC had a similar chemical shift to that observed at low resolution (3.2 +/- 0.2 ppm). This peak had a higher intensity in the xenografts than in normal human pancreatic tissue. HPLC analysis of the same lipid extracts from Capan-1 cells in culture, of tumors derived from these cells and from normal pancreas showed: 1. Identical concentrations of the different phospholipids from cancerous human pancreatic cells in vivo and in culture; and 2. A significantly higher level of PC in the extracts of normal human pancreatic tissue. The increase in intensity of the N(CH3)3 peak of PC in the Capan-1 tumors was not thought to be caused by an increase in PC concentration, but to a difference in conformation or mobility of the PC protons in the xenografts. The increase in relaxation time in cancerous tissue (from 60 to 125 ms) was also taken to be evidence in favor of a high mobility of protons.(ABSTRACT TRUNCATED AT 400 WORDS)     

Experimental allergic encephalomyelitis raises betaine levels in the spinal cord of strain 13 guinea-pigs

Chronic relapsing experimental allergic encephalomyelitis, an animal model of multiple sclerosis, was induced in Strain 13 guinea-pigs by subcutaneous injection of spinal cord homogenate and Freund's incomplete adjuvant supplemented with Mycobacterium tuberculosis. High resolution 1H NMR spectra of CNS tissue extracts indicated that the levels of choline metabolites, particularly betaine, were elevated in the spinal cord tissue, the principal site of lesion formation in this guinea-pig strain. The spectra also show that N-acetylated compounds are slightly depleted in the disease. The results are discussed in relation to the biochemical interpretation of NMR spectra obtained in vivo from patients with multiple sclerosis.     

Application of multilayer feed-forward neural networks to automated compound identification in low-resolution open-path FT-IR spectrometry

A drawback of current open-path Fourier transform infrared (OP/FT-IR) systems is that they need a human expert to determine those compounds that may be quantified from a given spectrum. In this work, multilayer feed-forward neural networks with one hidden layer were used to automatically recognize compounds in an OP/FT-IR spectrum without compensation of absorption lines due to atmospheric H2O and CO2. The networks were trained by fast-back-propagation. The training set comprised spectra that were synthesized by digitally adding randomly scaled reference spectra to actual open-path background spectra measured over a variety of path lengths and temperatures. The reference spectra of 109 compounds were used to synthesize the training spectra. Each neural network was trained to recognize only one compound in the presence of up to 10 other interferences in an OP/FT-IR spectrum. Every compound in a database of vaporphase reference spectra can be encoded in an independent neural network so that a neural network library can be established. When these networks are used for the identification of compounds, the process is analogous to spectral library searching. The effect of learning rate and band intensities on the convergence of network training was examined. The networks were successfully used to recognize five alcohols and two chlorinated compounds in field-measured controlled-release OP/FT-IR spectra of mixtures of these compounds.     

High-resolution mono- and multidimensional magic angle spinning 1H nuclear magnetic resonance of membrane peptides in nondeuterated lipid membranes and H2O

High-speed (14 kHz) solid-state magic angle spinning (MAS) 1H NMR has been applied to several membrane peptides incorporated into nondeuterated dilauroyl or dimyristoylphosphatidylcholine membranes suspended in H2O. It is shown that solvent suppression methods derived from solution NMR, such as presaturation or jump-return, can be used to reduce water resonance, even at relatively high water content. In addition, regioselective excitation of 1H peptide resonances promotes an efficient suppression of lipid resonances, even in cases where these are initially two orders of magnitude more intense. As a consequence, 1H MAS spectra of the peptide low-field region are obtained without interference from water and lipid signals. These display resonances from amide and other exchangeable 1H as well as from aromatic nonexchangeable 1H. The spectral resolution depends on the specific types of resonance and membrane peptide. For small amphiphilic or hydrophobic oligopeptides, resolution of most individual amide resonance is achieved, whereas for the transmembrane peptide gramicidin A, an unresolved amide spectrum is obtained. Partial resolution of aromatic 1H occurs in all cases. Multidimensional 1H-MAS spectra of membrane peptides can also be obtained by using water suppression and regioselective excitation. For gramicidin A, F2-regioselective 2D nuclear Overhauser effect spectroscopy (NOESY) spectra are dominated by intermolecular through-space connectivities between peptide aromatic or formyl 1H and lipid 1H. These appear to be compatible with the known structure and topography of the gramicidin pore. On the other hand, for the amphiphilic peptide leucine-enkephalin, F2-regioselective NOESY spectra mostly display cross-peaks originating from though-space proximities of amide or aromatic 1H with themselves and with aliphatic 1H. F3-regioselective 3D NOESY-NOESY spectra can be used to obtain through-space correlations within aliphatic 1H. Such intrapeptide proximities should allow determination of the conformation of the peptide in membranes. It is suggested that high-speed MAS multidimensional 1H NMR of peptides in nondeuterated membranes and in H2O can be used for studies of both peptide structure and lipid-peptide interactions.     

Tandem mass spectrometry and nuclear magnetic resonance spectroscopy studies of Candida bombicola sophorolipids and product formed on hydrolysis by cutinase

Natural mixtures of sophorolipids produced by the yeast Candida bombicola have been analyzed by fast atom bombardment (FAB)-MS and collision-induced dissociation (CID)-MS. Some pure components have been analysed by two-dimensional NMR spectroscopy. The presence of acidic, lactonic, and O-acetylated forms and the position of double bonds in the fatty acid part of these glycolipids can be easily inferred from positive and negative ion FAB-mass spectra. Details about position of O-acetylation can be obtained from CID mass spectra of [M+H]+ and [M-H]- ions and from the NMR spectra. Differences in CID fragmentation between protonated and sodiated molecular ions are discussed in detail. Enzymatic hydrolysis of 6',6"-di-O-acetyl sophorolipid lactone by cutinase from Fusarium solani results specifically in the removal of the 6'-O-acetyl group, whereas the 6"-O-acetyl and lactone group are resistant. This specificity is explained from a three-dimensional model of the sophorolipid generated on the basis of the short 1H,1H distances as inferred from the NMR (ROESY) spectra.     

Membrane structure and dynamics as viewed by solid-state NMR spectroscopy

The purpose of the present study is the investigation of the structure and dynamics of biological membranes using solid-state nuclear magnetic resonance (NMR) spectroscopy. Two approaches are used in our laboratory. The first involves the measurement of high-resolution 13C and 1H spectra obtained by the magic angle spinning (MAS) technique while the second approach involves the measurement of 31P and 2H powder spectra in static samples. This paper will present some recent results obtained by high-resolution solid-state 1H NMR on the conformation of gramicidin A incorporated in a phosphatidylcholine bilayers. More specifically, we were able to observe changes in the gramicidin spectra as a function of the cosolubilization solvent initially used to prepare the samples. The interaction between lipid bilayers and an anticancer drug derived from chloroethylurea was also investigated using proton NMR spectroscopy. Finally, we have studied the interaction between cardiotoxin, a toxic protein extracted from snake venom, and negatively charged lipid bilayers using 31P solid-state NMR spectroscopy.     

Investigation of germination and aging in Moravian III barley grain by nuclear magnetic resonance

High-resolution, solid-state 1H nuclear magnetic resonance (NMR) techniques are used for the first time to study germination in imbibed Moravian III barley grains. Whereas magic-angle spinning 1H NMR spectra reveal the water and lipid components in barley grains, combined rotation and multiple-pulse spectroscopy techniques provide 1H NMR spectra of grains that reveal the protein and carbohydrate as well as the water and lipid components. Spectra of grains are compared with spectra of model compounds to verify assignments. 1H T1 and T2 measurements using magic-angle spinning only and combined rotation and multiple-pulse spectroscopy techniques provide information about molecular mobility within the grains during inhibition. Some grains were subjected to artificial aging conditions. 1H NMR spectral comparisons are made between normal, viable grains and artificially aged grains.     

Neural network modeling of risk assessment in child protective services.

The advantages of using neural network methodology for the modeling of complex social science data are demonstrated, and neural network analysis is applied to Washington State Child Protective Services risk assessment data. Neural network modeling of the association between social worker overall assessment of risk and the 37 separate risk factors from the State of Washington Risk Assessment Matrix is shown to provide case classification results superior to linear or logistic multiple regression. The improvement in case prediction and classification accuracy is attributed to the superiority of neural networks for modeling nonlinear relationships between interacting variables; in this respect the mathematical framework of neural networks is a better approximation to the actual process of human decision making than linear, main effects regression. The implications of this modeling advantage for evaluating social science data within the framework of ecological theories are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Solid-state NMR studies of the prion protein H1 fragment

Conformational changes in the prion protein (PrP) seem to be responsible for prion diseases. We have used conformation-dependent chemical-shift measurements and rotational-resonance distance measurements to analyze the conformation of solid-state peptides lacking long-range order, corresponding to a region of PrP designated H1. This region is predicted to undergo a transformation of secondary structure in generating the infectious form of the protein. Solid-state NMR spectra of specifically 13C-enriched samples of H1, residues 109-122 (MKHMAGAAAAGAVV) of Syrian hamster PrP, have been acquired under cross-polarization and magic-angle spinning conditions. Samples lyophilized from 50% acetonitrile/50% water show chemical shifts characteristic of a beta-sheet conformation in the region corresponding to residues 112-121, whereas samples lyophilized from hexafluoroisopropanol display shifts indicative of alpha-helical secondary structure in the region corresponding to residues 113-117. Complete conversion to the helical conformation was not observed and conversion from alpha-helix back to beta-sheet, as inferred from the solid-state NMR spectra, occurred when samples were exposed to water. Rotational-resonance experiments were performed on seven doubly 13C-labeled H1 samples dried from water. Measured distances suggest that the peptide is in an extended, possibly beta-strand, conformation. These results are consistent with the experimental observation that PrP can exist in different conformational states and with structural predictions based on biological data and theoretical modeling that suggest that H1 may play a key role in the conformational transition involved in the development of prion diseases.     

Genetic Programming to Predict Bridge Pier Scour

Bridge-pier scour is a significant problem for the safety of bridges. Extensive laboratory and field studies have been conducted examining the effect of relevant variables. This note presents an alternative to the conventional regression-based equations (HEC-18 and regression equation developed by the writers), in the form of artificial neural networks (ANNs) and genetic programming (GP). There had been 398 data sets of field measurements that were collected from published literature and were used to train the network or evolve the program. The developed network and evolved programs were validated by using the observations that were not involved in the training. The performance of GP was found more effective when compared to regression equations and ANNs in predicting the scour depth at bridge piers.     

Biochemical effects of gadolinium chloride in rats liver and kidney studied by ^1H NMR metabolomics

The biochemical effects of gadolinium chloride were studied using high-resolution IH nuclear magnetic resonance (NMR) spec-troscopy to investigate the biochemical composition of tissue (liver and kidney) aqueous extracts obtained from control and gadolinium chlo-ride (GdCl3) (10 and 50 mg/kg body weight, intraperitoneal injection, i.p.) treated rats. Tissue samples were collected at 48, 96 and 168 h p.d. after exposure to GdCl3, and extracted using methanol/chloroform solvent system. 1H NMR spectra of tissue extracts were analyzed by pat-tern recognition using principal components analysis. The liver damages caused by GdCl3 were characterized by increased succinate and de-creased glycogen level and elevated lactate, alanine and betaine concentration in liver. Furthermore, the increase of creatine and lactate, and decrease of glutamate, alanine, phosphocholine, glycophosphocholine (GPC), betaine, myo-inositoi and trimethylamine N-oxide (TMAO)levels in kidney illustrated kidney disturbance induced by GdCl3.     

Classification of low back pain from dynamic motion characteristics using an artificial neural network

STUDY DESIGN: Data were collected from 183 subjects who were randomly assigned to the training and test groups. During testing of the classification system, knowledge of the low back pain condition or motion characteristics of the patients in the test group was not made available to the system. OBJECTIVES: To determine specific characteristics of trunk motion associated with different categories of spinal disorders and to determine whether a neural network analysis system can be effective in distinguishing patterns. SUMMARY OF BACKGROUND DATA: Numerous studies have established the difficulty of evaluating lower back pain. Imaging techniques are expensive and ineffective in many cases. A technique for evaluation of lower back pain was developed on the basis of analysis of such dynamic motion features as shape, velocity, and symmetry of movements, using a neural network classification system. METHODS: Dynamic motion data were collected from 183 subjects using a triaxial goniometer. Features of the movement were extracted and provided as input to a two-stage neural network classifier governed by a radial basis function architecture. After training, the output of the classifier was compared with Québec Task Force pain classifications obtained for the patients. Linear and nonlinear classification techniques were compared. RESULTS: The system could determine low back pain classification from motion characteristics. The neural network classifier produced the best results with up to 85% accuracy on novel "validation" data. CONCLUSIONS: A neural network based on kinematic data is an excellent predictive model for classification of lower back pain. Such a system could markedly improve the management of lower back pain in the individual patient.     

基于神经网络集成-X射线荧光光谱法的铁矿石中全铁含量测定

为了探索人工智能在铁矿石品质快速检验中的应用,研究了机器学习算法与化学计量学和X射线荧光光谱仪(XRF)相结合快速测定铁矿石中全铁含量的方法。收集来自于不同产地的,主要物相为赤铁矿、褐铁矿、磁铁矿、针铁矿和多物相混和结构的铁矿石样品共1098个作为样本集。采用X射线荧光光谱仪对铁矿石样品熔片进行扫描,扫描后的光谱图提取数据点后作为神经网络的输入,以全铁含量作为输出结果。然后依据X射线衍射(XRD)得到的物相结构优化自组织(SOM)网络,并对全部样本的XRF图谱进行分类,对分类后的每一个子集分别采用反向传播(BP)和径向基函数(RBF)网络建立回归子模型,对各子模型的预测结果进行整合,最终建立基于集成神经网络和X射线荧光光谱法的铁矿石中全铁含量预测模型。方法模型建立后,不需要额外标准物质建立校准曲线,能够实现对未知样品的分类和输出全铁含量结果。     

Proton nuclear magnetic resonance studies on dideoxyribonucleoside methylphosphonates

A series of dideoxyribonucleoside methylphosphonates, d-ApA, d-ApT, d-TpA, and TpT, were synthesized chemically and the diastereoisomers of each dimer were separated [Miller, P. S., Yano, J., Yano, E., Carroll, C., Jayaraman, K., & Ts'o, P. O. P. (1979) Biochemistry 18, 5134]. The 1H NMR spectra of these compounds are similar to those of their parent diester compounds. Specifically, the assignments of the 1H resonances of the two diastereoisomers of d-ApA (designated as 1 and 2) were reaffirmed by comparing with the unmodified, parent d-ApA. The absolute configuration of the phosphonate methyl group of the two isomers (d-ApA)1 and (d-ApA)2 was determined by the NOE technique. The 1H NMR spectra of the diastereoisomers of d-ApA, as well as the corresponding monomer components dAp and CH3pdA, and TpT were analyzed by spectrum simulation techniques. Thus, all the coupling constants and chemical shifts of the proton resonances of the deoxyribofuranose ring and the phosphonate methyl group could be precisely determined. These data provide the information for an analysis of the sugar puckering and backbone conformations of these novel nonionic nucleic acid analogues. It was found that the conformations of the sugar-phosphate backbones of each isomer are similar to each other and are similar to the conformations of the parent dinucleoside monophosphates. The average adenine stacking conformations of (d-ApA)1 and (d-ApA)2 were described in numerical coordinates derived from a computer analysis which included both ring-current magnetic anisotropy and atomic diamagnetic anisotropy effects. The two computer-derived conformational models are similar to those derived from the graphic approximation based only on the ring-current effects. For each pair of dimer analogues, the base stacking mode of isomer 1 is similar to that of its parent diester while the extent of base overlap in isomer 2 is less than that in isomer 1. The results of the conformational analysis based on NMR data are consistent with the results obtained from ultraviolet and circular dichroism measurements on these dimers.     

1H/27Al TRAPDOR NMR studies on aluminum species in dealuminated zeolites

Aluminum species in several dealuminated zeolites (ultrastable HY, HZSM-5 and mordenite) were investigated in detail by means of the newly introduced 1H/27Al TRAPDOR method in combination with 27Al MAS NMR, and the quadrupole coupling constants (Q[CC]s) for aluminum atoms associated with these species were obtained. A signal at ca. 6.8 ppm, due to water molecules adsorbed on Lewis acid sites, was observed in the 1H MAS spectra for all the three zeolites. The TRAPDOR NMR provides direct evidence that there is a strong interaction between the adsorbed water molecules and the aluminum atoms of the Lewis-acid sites. The Q(CC) values for this aluminum species of 8.3, 6.7 and 11.3 MHz were determined from the TRAPDOR profiles for the ultrastable HY, HZSM-5 and mordenite zeolites, respectively. The Q(CC)s calculated from the TRAPDOR curves are usually larger than 10 MHz for both Bronsted-acid sites (SiOHAI) and non-framework aluminum species in the three zeolites. Three narrow peaks at 54, 30 and 0 ppm are separately superimposed on a broad hump in the 27Al MAS spectra of the three dehydrated zeolites, while the latter is associated with the 'NMR invisible' Al. The NMR experimental results suggest that the three kinds of aluminum species (non-framework aluminum species, Bronsted- and Lewis-acid sites) are all responsible for the resonance of the broad hump in dehydrated zeolites, which makes it difficult to explain the 27Al MAS spectra. Fortunately, the TRAPDOR NMR provides a direct method for individually studying different aluminum species with large Q(CC)s via their dipolar coupling to nearby proton nuclei.