Constrained learning vector quantization

Kohonen's learning vector quantization (LVQ) is an efficient neural network based technique for pattern recognition. The performance of the method depends on proper selection of the learning parameters. Over-training may cause a degradation in recognition rate of the final classifier. In this paper we introduce constrained learning vector quantization (CLVQ). In this method the updated coefficients in each iteration are accepted only if the recognition performance of the classifier after updating is not decreased for the training samples compared with that before updating, a constraint widely used in many prototype editing procedures to simplify and optimize a nearest neighbor classifier (NNC). An efficient computer algorithm is developed to implement this constraint. The method is verified with experimental results. It is shown that CLVQ outperforms and may even require much less training time than LVQ.

Relationship of the celiac and superior mesenteric arteries to the vertebral bodies and its clinical relevance

To determine the relationship of the origins of the celiac artery (CA) and superior mesenteric artery (SMA) to the vertebral bodies, and their distribution in males and females, we examined them quantitatively on the angiograms of 1,591 patients (male 1,003, female 588). Our results showed that the origins of the CA and SMA were highly variable: 14.0% of the CA originated beyond the extent of the lower half of T12 to the upper part of L1, and 24.7% of the SMA originated beyond the level of L1. The origin of the SMA was more variable than that of the CA (p < 0.01). The variation in the origin of the CA was not significantly different (p > 0.05) between males (13.4%) and females (15.1%), but a statistically significant difference (p < 0.01) between males (22.3%) and females (28.7%) was shown for the SMA. The distance between the origin of the two arteries was over 3 cm in six patients (0.4%). Our results provide a valuable reference for clinical practice.

Functional consequences of iron supplementation in iron-deficient female cotton mill workers in Beijing, China

Eighty iron-deficient, nonpregnant female workers were randomly assigned to ferrous sulphate (60 or 120 mg Fe/d) or placebo treatment for 12 wk. Energy expenditure was estimated during 3 d by heart rate (HR) recording. Production efficiency (PE) was calculated as the ratio of productivity to energy expenditure. In the iron-treated group mean hemoglobin (Hb) increased from 114 to 127 g/L (P < 0.001), mean serum ferritin increased from 9.7 to 30.0 micrograms/L (P < 0.001), and mean free erythrocyte protoporphyrin decreased from 1.01 to 0.49 mumol/L (P < 0.001). Mean HR at work decreased from 95.5 to 91.1 beats/min (P < 0.001), which was inversely correlated with the change in Hb (r = -0.60, P < 0.001). PE increased significantly in the iron-treated group (P < 0.001) and its change paralleled the change in Hb (r = 0.58, P < 0.001). The results show that iron supplementation enabled these women to do the same work at a lower energy cost.

p135tyk2, an interferon-alpha-activated tyrosine kinase, is physically associated with an interferon-alpha receptor

Recent genetic studies have linked the tyk2 gene, which encodes a novel type of non-receptor tyrosine kinase, to the interferon-alpha intracellular signaling pathway. In this report, biochemical evidence is presented which supports this proposed function for the tyk2 tyrosine kinase and further defines its role in the interferon-alpha signaling cascade. Specifically, the tyk2 gene is shown to encode a 135-kDa protein which is rapidly phosphorylated on tyrosine in response to interferon-alpha treatment. Indirect evidence suggests that the tyrosine phosphorylation of p135tyk2 is the result of autokinase activity, implying that the Tyk2 tyrosine kinase is activated by interferon-alpha treatment. Two complementary methods demonstrate a physical association between p135tyk2 and the alpha-subunit of the interferon-alpha receptor. First, immunoblots show that monoclonal antibodies against the alpha-subunit of the interferon-alpha receptor can co-immunoprecipitate p135tyk2. Second, interferon-alpha receptor proteins which have been labeled by affinity cross-linking with 125I-interferon-alpha 2 can be co-immunoprecipitated using anti-tyk2 antisera. Taken together, these data suggest that an interferon-alpha receptor-p135tyk2 complex functions, in a manner analogous to the CD4-lck tyrosine kinase complex, to initiate the interferon-alpha signaling cascade.

Interaction of calmodulin with synthetic deletion peptides of melittin

The 26-residue peptide melittin present in bee venom has been shown to bind calmodulin tightly. In this study we synthesized the following series of deletion peptides of melittin by the solid-phase method: Mel12, Mel13, Mel14, Mel15, Mel15F. The results of this study show that the deletion peptides Mel14 and Mel15 have almost the same binding activity as the intact native peptide. Each deletion peptide forms a 1:1 complex with calmodulin according to electrophoresis analysis. When the tryptophanyl residue of Mel15 was replaced by the phenylalaninyl residue, the dissociation constant of the peptide-calmodulin complex increased. This shows the importance of the tryptophanyl residue for binding to calmodulin.

A novel receptor tyrosine phosphatase-sigma that is highly expressed in the nervous system

A novel transmembrane receptor protein tyrosine phosphatase-sigma (RPTP-sigma) was cloned from a rat brain stem cDNA library. The extracellular segment of one form of RPTP-sigma contains 824 amino acids and is composed of three immunoglobulin-like and five fibronectin type III (FNIII)-like repeats. The 627-amino acid cytoplasmic region of RPTP-sigma consists of two catalytic domains oriented in tandem. Northern blot analyses indicate that RPTP-sigma is highly expressed in the brain as two major transcripts of 5.7 and 6.9 kilobases (kb). The 5.7-kb transcript is expressed exclusively in the brain while the 6.9-kb species can be detected in the lung and heart, but at significantly lower levels. In situ hybridization studies confirm that RPTP-sigma is localized predominantly in the nervous system and can be detected in the rat as early as embryonic day 12. During embryonic development, RPTP-sigma is expressed extensively in the central and peripheral nervous systems, including the trigeminal and dorsal root ganglia as well as the retina. In adult rat brain, expression is restricted primarily to the olfactory tubercule, cerebellum, and hippocampus. Within the latter structure, RPTP-sigma is present in the pyramidal cell layer and granular layer of the dentate gyrus. Transfection of RPTP-sigma cDNA into human embryonic kidney 293 cells results in the synthesis of a protein with an apparent molecular mass of 200 kDa as detected by immunoprecipitation and immunoblot analyses using polyclonal antibodies against the FNIII-like repeats present in the extracellular domain of RPTP-sigma. The gene for RPTP-sigma has been mapped to distal chromosome 17 in the mouse.

Mechanism of adenylate kinase. 1H, 13C, and 15N NMR assignments, secondary structures, and substrate binding sites

Backbone 1H, 13C, and 15N NMR assignments were obtained for the complex of chicken muscle adenylate kinase (AK) with its bisubstrate analog, MgAP5A [magnesium P1,P5-bis(5'-adenosyl)-pentaphosphate]. The assignments were used to elucidate the secondary structures and the enzyme-MgAP5A interactions. The work involves two unusual features: the molecular weight of AK (21.6 kDa) is one of the largest, on a monomeric basis, for which nearly complete assignment has been reported to date, and the assignment was performed at pH 7.1 instead of the acidic pH used for most other proteins. The results are summarized as follows. Firstly, unambiguous sequential assignments of backbone resonances have been achieved effectively by the combined use of two sequential assignment methods: NOE-directed assignments and the recently developed 1J-coupling-directed assignments. The starting points of the assignments were provided by several specifically labeled enzyme samples. Over 90% of the backbone 1H, 13C, and 15N resonances have been assigned. Secondly, spin system information was obtained from the HCCH-TOCSY and HCCH-COSY experiments as well as from 2D homonuclear NMR data. Overall, the side-chain resonances of ca. 40% of the residues, including most of the those displaying NOEs with the adenosine moieties of MgAP5A, have been assigned. Thirdly, secondary structural elements in the AK-MgAP5A complex were identified by extensive analyses of 1H-15N 2D HMQC-NOESY and 3D NOESY-HMQC spectra. Overall, the enzyme consists of ca. 60% alpha-helices and a five-stranded parallel beta-sheet. The results are compared with the secondary structure of the free AK from porcine muscle in crystals [Dreusicke, D., Karplus, P. A., & Schulz, G. E. (1988) J. Mol. Biol. 199, 359-371]. Lastly, most of the intermolecular NOEs between AK and the adenosine moieties of MgAP5A have been identified: Thr39, Leu43, Gly64, Leu66, Val67, Val72, and Gln101 are in proximity to the adenosine moiety of the adenosine 5'-monophosphate site, whereas Thr23 is in proximity to that of the adenosine 5'-triphosphate site. These data are discussed in relation to previous results from site-directed mutagenesis, NMR, and X-ray studies and in relation to the mechanism of catalysis.

Cell cycle-regulated nuclear localization of MCM2 and MCM3, which are required for the initiation of DNA synthesis at chromosomal replication origins in yeast

MCM2 and MCM3 are two genetically interacting and structurally related proteins essential for growth in Saccharomyces cerevisiae. Mutants defective in these proteins affect the stability of minichromosomes in general, but the severity of the defect is dependent on the autonomously replicating sequence (ARS) that drives the replication of that plasmid. In this paper we show by two-dimensional gel electrophoresis that the initiation of DNA synthesis at chromosomal replication origins is also reduced in frequency in these mutants. We show further that the nuclear and subnuclear localizations of the MCM2 and MCM3 proteins are temporally regulated with respect to the cell cycle. These proteins enter the nucleus at the end of mitosis, persist there throughout G1 phase, and disappear from it at the beginning of S phase. Once inside the nucleus, a fraction of the MCM2 and MCM3 proteins becomes tightly associated with DNA. The association of MCM2 and MCM3 with chromatin presumably leads to the initiation of DNA synthesis, and their subsequent disappearance from the nucleus presumably prevents reinitiation of DNA synthesis at replication origins. This temporally and spatially restricted localization of MCM2 and MCM3 in the nucleus may serve to ensure that DNA replication occurs once and only once per cell cycle.

DNA fingerprinting in relation to male dominance and paternity in a captive colony of tammar wallabies (Macropus eugenii)

The tammar wallaby has a polygynous mating system in which the dominant male usually controls initial access to oestrous females by mating first and then guarding the female from the advances of other subordinate males. In this study we used DNA fingerprinting with a human 3' hypervariable region (HVR) alpha globin probe to examine the paternity of pouch young progeny from 13 female tammars that were given continual access during the breeding season to a group of four sexually mature males. Constant individual-specific DNA profiles were observed for each animal. Paternity for 22 pouch young was successfully assigned using visual and computer-based analyses. However, no statistical difference was observed between the number of young sired by any of the four males (chi 2 = 2, d.f. = 3, P > 0.1). Mate guarding by the dominant male in our captive breeding group was not, therefore, sufficient to prevent successful subsequent matings by subordinates nor to enhance the genetic contribution of this male to the next generation. In each analysis, visual and computer assignments of paternity coincided, and these concurred with the results of a relatedness test which found that a large number of DNA bands were shared by sires and their progeny. The results from this paternity study show that first mating and subsequent mate guarding by the dominant male tammar wallaby in our captive group do not significantly skew the outcome of paternity towards this male and away from other males that subsequently mate with each female.

Fat tissue and fat suppression

Fat tissues consist of fat cells, capillaries, and collagen fibers. In order to completely suppress the signals from fat tissues in clinical magnetic resonance imaging, the signal from capillaries and collagen fibers as well as from fat cells should all be suppressed. We have previously reported that fat signal can be uniformly suppressed by applying an optimized presaturation pulse. The inhomogeneously broadened fat peak of tissue spectrum is excited by the optimized pulse and dephased by a subsequent field gradient. The broadened water peak is not affected. In this paper we discuss a technique that suppresses signals from fat tissues completely as well as uniformly. This technique is based on the cancellation of fat and water signals in the same image voxel by combining the optimized selective excitation with the opposite phase imaging technique. Experimental and clinical images demonstrate that the new technique improves the delineation and depiction of anatomy in clinical fat suppression imaging.

Data truncation artifact reduction in MR imaging using a multilayer neural network

A magnetic resonance image (MRI) may contain truncation artifacts if there are not enough high-frequency data when the conventional Fourier transform method is used for reconstruction. A method for reducing the artifacts using a multilayer neural network is presented. The network consists of one linear output layer and at least one nonlinear hidden layer. The missing high-frequency components are predicted based on known low-frequency components and are used to reduce the truncation artifacts of the image. Results from a series of simulation experiments are discussed.

MR image compression using iterated function systems

This paper presents a method for MR image compression using iterated function systems based on fractal models. The method makes use of self-similarities of an image and achieves compression by matching similar image blocks. A fast matching procedure is proposed in this paper for efficient image encoding. The method is tested with real MR images.

Beneficial effects of berberine on hemodynamics during acute ischemic left ventricular failure in dogs

In 18 dogs ischemic left ventricular failure characterized by a 30 percent reduction in peak rate of rise of left ventricular pressure (+dp/dt) and elevation of left ventricular end-diastolic pressure (LVEDP) to 15 mmHg or more was produced by ligation of the proximal left anterior descending coronary artery followed by serial occlusions of the distal left circumflex coronary artery. In 10 days, administration of berberine in an intravenous bolus injection (1 mg/kg, within 3 minutes) followed by a constant infusion (0.2 mg/kg/min, 30 minutes) increased the cardiac output (CO) from 1.25 +/- 0.12 to 1.61 +/- 0.17 L/min (P < 0.05), and +dp/dt from 810 +/- 85 to 1021 +/- 130 mmHg/s (P < 0.01), and decreased LVEDP from 16.5 +/- 1.3 to 12.0 +/- 1.0 mmHg (P < 0.05), diastolic blood pressure from 94 +/- 6 to 84 +/- 5 mmHg (P < 0.01), systemic vascular resistance from 7303 +/- 278 to 5442 +/- 231 dynes.x/cm5 (P < 0.01), but did not affect the heart rate. Injection of 5% glucose with the same volume did not improve CO and dp/dt (P > 0.05) but increased the LVEDP from 17.1 +/- 1.4 to 17.8 +/- 1.6 mmHg (P < 0.01) in 8 dogs. The levels of plasma concentration of berberine was determined with high-performance liquid chromatography. The changes in plasma drug level were found parallel to hemodynamic effects of berberine. The results of this study showed that berberine was able to improve the impaired left ventricular function by its positive inotropic effect and mild systemic vasodilatation.

Density-dependent modulation of vascular smooth muscle alpha-actin biosynthetic processing in differentiated BC3H1 myogenic cells

The expression of vascular smooth muscle (VSM) alpha-actin mRNA during BC3H1 myogenic cell differentiation is specifically stimulated by conditions of high cell density. Non-proteolytic dissociation of cell-cell and cell-matrix contacts in post-confluent cultures of BC3H1 myocytes using EDTA promotes loss of the differentiated morphological phenotype. EDTA-dispersed myocytes exhibit an undifferentiated fibroblastoid appearance and contained reduced levels of both VSM and skeletal alpha-actin mRNA. Muscle alpha-actin mRNA levels in EDTA-dispersed myocytes were not restored to that observed in confluent myocyte preparations by experimental manipulation of cell density conditions. Pulse-labeling techniques using L-[35S]cysteine to identify muscle actin biosynthetic intermediates revealed that EDTA-dispersed myocytes expressed nascent forms of both the VSM and skeletal muscle alpha-actin polypeptide chains. However EDTA-dispersed myocytes were less efficient in the post-translational processing of immature VSM alpha-actin compared to non-dispersed myocytes. Simple cell-to-cell contact may mediate VSM alpha-actin processing efficiency since high-density preparations of EDTA-dispersed myocytes processed more VSM alpha-actin intermediate than myocytes plated at low density. The actin isoform selectivity of the response to modulation of intercellular contacts suggests that actin biosynthesis in BC3H1 myogenic cells involves mechanisms capable of discriminating between different isoform classes of nascent actin polypeptide chains.

An efficient algorithm for MR image reconstruction and compression

In magnetic resonance imaging (MRI), the original data are sampled in the spatial frequency domain. The sampled data thus constitute a set of discrete Fourier transform (DFT) coefficients. The image is usually reconstructed by taking inverse DFT. The image data may then be efficiently compressed using the discrete cosine transform (DCT). We present here a method of using DCT to treat the sampled data, which combines two procedures, image reconstruction and data compression. This method may be particularly useful in medical picture archiving and communication systems (PACS) where both image reconstruction and compression are important issues.

Mechanism of adenylate kinase. Structural and functional roles of the conserved arginine-97 and arginine-132

The structural and functional roles of two conserved active site residues, Arg-97 and Arg-132, in chicken muscle adenylate kinase (AK) were evaluated by site-directed mutagenesis in conjunction with one- and two-dimensional proton nuclear magnetic resonance (NMR), kinetics, and guanidine hydrochloride-induced denaturation. In addition, 31P NMR analysis was used to evaluate the contribution of Arg-97 to the phosphorus stereospecificity of AK. The results and conclusions are summarized as follows: (i) Kinetic analysis of R97M reveals 6- and 28-fold increases in the dissociation constant Ki and Michaelis constant K of AMP, respectively, and a moderate 30-fold decrease in kcat. The Ki and K values of MgATP are relatively unperturbed. The localized effect of AMP stabilization was independently confirmed by proton NMR titration, which showed a ca. 20-fold increase in the dissociation constant of AMP but not of MgATP. (ii) R132M affords a dramatic decrease in kcat by a factor of 8.0 x 10(3), with unchanged dissociation and Michaelis constants for either substrate. The lack of perturbation in the affinities toward substrates was confirmed by proton NMR titration. (iii) Although small chemical shift changes were observed for the free mutants and their complexes with substrates, further analyses by nuclear Overhauser enhanced spectroscopy with the bisubstrate analogue inhibitor, P1,P5-bis(5'-adenosyl)pentaphosphate (AP5A), indicated little perturbation in the global conformation. (iv) Contributions to conformational stability by Arg-97 and Arg-132 are negligible on the basis of the free energy of unfolding, delta GdH2O. (v) R97M was predicted and demonstrated to exhibit enhanced stereospecificity at the AMP site by at least 10-fold relative to WT in the conversion of adenosine 5'-monothiophosphate to adenosine 5'-(1-thiodiphosphate). This result for R97M was predicted on the basis of the orientation of Arg-97 relative to Arg-44 and AMP in the active site as observed in available crystal structures and the stereospecificity results of R44M [Jiang, R.-T., Dahnke, T., & Tsai, M.-D. (1991) J. Am. Chem. Soc. 113, 5485-5486]. (vi) The above structural and functional analyses led us to conclude that Arg-97 interacts with the phosphoryl group of AMP, beginning at the binary complex (1-2 kcal/mol), continuing through the transition state (3.5 kcal/mol), and that Arg-132 stabilizes the transition state by greater than 5 kcal/mol. (vii) The functional importance of Arg-97 appears to be similar to that of Arg-44 [Yan, H., Dahnke, T., Zhou, B., Nakazawa, A., & Tsai, M.-D. (1990) Biochemistry 29, 10956-10964].(ABSTRACT TRUNCATED AT 400 WORDS)

[Effect of hypoxia on vasodilation induced by substance P (SP) and calcitonin gene-related peptide (CGRP)]

In this paper, isolated rabbit arteries (renal artery, RA; femoral artery, FA; basilar artery, BA) were used to study the effect of hypoxia on vasodilation induced by substance P (SP) and calcitonin gene-related peptide (CGRP). The results showed that hypoxia caused a marked decrease of the arteries, response to SP (10(-7) mol/L) and CGRP (2.8 x 10(-9) mol/L). Repeated hypoxia even resulted in a complete disappearance of response. On the second hypoxia, CGRP (2.8 x 10(-9) mol/L) induced vasodilation on BA was vanished. On the third hypoxia, vasodilation induced by CGRP (2.8 x 10(-9) mol/L) on FA and that induced by SP (2 x 10(-7) mol/L) on BA were vanished, and CGRP (2.8 x 10(-9) mol/L) induced a lesser extent of vasodilation on RA. At the fourth time, SP (10(-7) mol/L) -induced vasodilation of FA disappeared, and that on RA was small. At the fifth time, the effect of SP (10(-7) mol/L) on RA was nearly vanished. It is concluded that the effect of hypoxia on CGRP-induced vasodilation is stronger than that on SP-induced vasodilation, and the effect is most pronounced on the BA.

Fat suppression with an improved selective presaturation pulse

Chemical shift selective (CHESS) imaging is an efficient and easily implemented technique for suppression of the fat component in clinical images. A Gaussian pulse is widely used as the selective presaturation pulse in the CHESS technique. However, a Gaussian pulse hardly performs well even if the magnetic field is very carefully shimmed. In clinical applications, the fat and water peaks are always inhomogeneously broadened. We have analyzed the performance of the Gaussian pulse in the presence of magnetic field inhomogeneities and present a more efficient selective presaturation pulse developed by applying the conjugate gradient method. The resulting pulse performs well without any requirement for magnetic field homogeneity greater than needed for routine diagnostic MR imaging.

Motion artifact suppression: a review of post-processing techniques

Patient motion during data acquisition in magnetic resonance imaging causes artifacts in the reconstructed image, which for two-dimensional Fourier transform imaging techniques appear as blurring and ghost repetitions of the moving structures. While the problem with intra-view effects has been effectively addressed using gradient moment nulling techniques, there is no corresponding technique for inter-view effects with equal effectiveness and general applicability. A number of techniques have been proposed for correcting the inter-view effects, and these may be divided into those that minimise the corruption of the data, and those that post-process the data to restore the image. The techniques in the former category are briefly reviewed, then those in the latter category are examined in detail. These are analysed in terms of motion model, model parameter estimation, and data correction.