Efficient Optimization Algorithms for Nonlinear Data Analysis

Identification of low-dimensional structures and main sources of variation from multivariate data are fundamental tasks in data analysis. Many methods aimed at these tasks involve solution of an optimization problem. Thus, the objective of this thesis is to develop computationally efficient and theoretically justified methods for solving such problems. Most of the thesis is based on a statistical model, where ridges of the density estimated from the data are considered as relevant features. Finding ridges, that are generalized maxima, necessitates development of advanced optimization methods. An efficient and convergent trust region Newton method for projecting a point onto a ridge of the underlying density is developed for this purpose. The method is utilized in a differential equation-based approach for tracing ridges and computing projection coordinates along them. The density estimation is done nonparametrically by using Gaussian kernels. This allows application of ridge-based methods with only mild assumptions on the underlying structure of the data. The statistical model and the ridge finding methods are adapted to two different applications. The first one is extraction of curvilinear structures from noisy data mixed with background clutter. The second one is a novel nonlinear generalization of principal component analysis (PCA) and its extension to time series data. The methods have a wide range of potential applications, where most of the earlier approaches are inadequate. Examples include identification of faults from seismic data and identification of filaments from cosmological data. Applicability of the nonlinear PCA to climate analysis and reconstruction of periodic patterns from noisy time series data are also demonstrated. Other contributions of the thesis include development of an efficient semidefinite optimization method for embedding graphs into the Euclidean space. The method produces structure-preserving embeddings that maximize interpoint distances. It is primarily developed for dimensionality reduction, but has also potential applications in graph theory and various areas of physics, chemistry and engineering. Asymptotic behaviour of ridges and maxima of Gaussian kernel densities is also investigated when the kernel bandwidth approaches infinity. The results are applied to the nonlinear PCA and to finding significant maxima of such densities, which is a typical problem in visual object tracking.

Differential effects of lead and zinc on inhibitory avoidance learning in mice

We studied the effects of chronic intoxication with the heavy metals lead (Pb2+) and zinc (Zn2+) on memory formation in mice. Animals were intoxicated through drinking water during the pre- and postnatal periods and then tested in the step-through inhibitory avoidance memory task. Chronic postnatal intoxication with Pb2+ did not change the step-through latency values recorded during the 4 weeks of the test (ANOVA, P>0.05). In contrast, mice intoxicated during the prenatal period showed significantly reduced latency values when compared to the control group (day 1: q = 4.62, P<0.05; day 7: q = 4.42, P<0.05; day 14: q = 5.65, P<0.05; day 21: q = 3.96, P<0.05, and day 28: q = 6.09, P<0.05). Although chronic postnatal intoxication with Zn2+ did not alter a memory retention test performed 24 h after training, we noticed a gradual decrease in latency at subsequent 4-week intervals (F = 3.07, P<0.05), an effect that was not observed in the control or in the Pb2+-treated groups. These results suggest an impairment of memory formation by Pb2+ when the animals are exposed during the critical period of neurogenesis, while Zn2+ appears to facilitate learning extinction.

Effects of lipopolysaccharide on low- and high-density cultured rabbit vascular smooth muscle cells: differential modulation of nitric oxide release, ERK1/ERK2 MAP kinase activity, protein tyrosine phosphatase activity, and DNA synthesis

Previous studies have shown that exogenously generated nitric oxide (NO) inhibits smooth muscle cell proliferation. In the present study, we stimulated rabbit vascular smooth muscle cells (RVSMC) with E. coli lipopolysaccharide (LPS), a known inducer of NO synthase transcription, and established a connection between endogenous NO, phosphorylation/dephosphorylation-mediated signaling pathways, and DNA synthesis. Non-confluent RVSMC were cultured with 0, 5, 10, or 100 ng/ml of the endotoxin. NO release was increased by 86.6% (maximum effect) in low-density cell cultures stimulated with 10 ng/ml LPS as compared to non-stimulated controls. Conversely, LPS (5 to 100 ng/ml) did not lead to enhanced NO production in multilayered (high density) RVSMC. DNA synthesis measured by thymidine incorporation showed that LPS was mitogenic only to non-confluent RVSMC; furthermore, the effect was prevented statistically by aminoguanidine (AG), a potent inhibitor of the inducible NO synthase, and oxyhemoglobin, an NO scavenger. Finally, there was a cell density-dependent LPS effect on protein tyrosine phosphatase (PTP) and ERK1/ERK2 mitogen-activated protein (MAP) kinase activities. Short-term transient stimulation of ERK1/ERK2 MAP kinases was maximal at 12 min in non-confluent RVSMC and was prevented by preincubation with AG, whereas PTP activities were inhibited in these cells after 24-h LPS stimulation. Conversely, no significant LPS-mediated changes in kinase or phosphatase activities were observed in high-density cells. LPS-induced NO generation by RVSMC may switch on a cell density-dependent proliferative signaling cascade, which involves the participation of PTP and the ERK1/ERK2 MAP kinases.

Differential expression of AMPA-type glutamate receptor subunits during development of the chick optic tectum

Glutamate receptors have been often associated with developmental processes. We used immunohistochemical techniques to evaluate the expression of the AMPA-type glutamate receptor (GluR) subunits in the chick optic tectum (TeO). Chick embryos from the 5th through the 20th embryonic day (E5-E20) and one-day-old (P1) chicks were used. The three types of immunoreactivity evaluated (GluR1, GluR2/3, and GluR4) had different temporal and spatial expression patterns in the several layers of the TeO. The GluR1 subunit first appeared as moderate staining on E7 and then increased on E9. The mature GluR1 pattern included intense staining only in layer 5 of the TeO. The GluR2/3 subunits presented low expression on E5, which became intense on E7. The staining for GluR2/3 changed to very intense on E14 in tectal layer 13. Staining of layer 13 neurons is the most prominent feature of GluR immunoreactivity in the adult TeO. The GluR4 subunit generally presented the lowest expression starting on E7, which was similar to the adult pattern. Some instances of transient expression of GluR subunits were observed in specific cell populations from E9 through E20. These results demonstrate a differential expression of the GluR subunits in the embryonic TeO, adding information about their possible functions in the developmental processes of the visual system.

A brain microdialysis study on 5-HT release in freely moving rat lines selectively bred for differential 5-HT1A receptor function

Breeding for high and low hypothermic responses to systemic administration of a serotonin1A (5-HT1A) receptor agonist (8-hydroxy-2-(di-n-propylamino)tetralin, 8-OH-DPAT) has resulted in high DPAT-sensitive (HDS) and low DPAT-sensitive (LDS) lines of rats, respectively. These lines also differ in several behavioral measures associated with stress. In the present microdialysis study we observed that basal 5-HT concentrations in the prefrontal cortex and dorsal hippocampus did not differ significantly between HDS and LDS rats. Thus, behavioral differences between the HDS and LDS lines might not be attributed to differences in basal 5-HT release. However, both lines had lower basal levels of 5-HT release than their randomly bred control group (random DPAT-sensitive, RDS) in the prefrontal cortex (mean ± SEM, pg/20 µl, was 3.0 ± 0.4 for LDS, 3.8 ± 0.3 for HDS and 6.4 ± 0.6 for RDS; F(2,59) = 5.8, P<0.005). The administration of (±)-fenfluramine (10 mg/kg) induced a greater increase in hippocampal 5-HT levels in HDS rats (500%) as compared with LDS (248%) or RDS (243%) rats (P<0.0001). There were no significant differences in the prefrontal cortex among lines, with a fenfluramine-induced 5-HT increase of about 900% in the three groups. This differential response to fenfluramine may be due to functional alterations of hippocampal 5-HT reuptake sites in the HDS line.

Differential expression of integrin subunits on adherent and nonadherent mast cells

Mast cell progenitors arise in bone marrow and then migrate to peripheral tissues where they mature. It is presumed that integrin receptors are involved in their migration and homing. In the present study, the expression of various integrin subunits was investigated in three systems of adherent and nonadherent mast cells. Mesentery mast cells, freshly isolated bone marrow-derived mast cells (BMMC) and RBL-2H3 cells grown attached to tissue culture flasks are all adherent mast cells and peritoneal mast cells, and cultured BMMC and RBL-2H3 cells grown in suspension represent nonadherent mast cell populations. Pure populations of mast cells were immunomagnetically isolated from bone marrow, mesentery and peritoneal lavage using the mast cell-specific monoclonal antibody AA4. By immunomicroscopy, we could demonstrate that all of these mast cells expressed alpha4, alpha5, alpha6, ß1 and ß7 integrin subunits. The expression of the alpha4 integrin subunit was 25% higher in freshly isolated mesentery mast cells and BMMC. Consistent with the results obtained by immunomicroscopy, mesentery mast cells expressed 65% more mRNA for the alpha4 integrin subunit than peritoneal mast cells. In vitro studies were also conducted using the rat mast cell line RBL-2H3. RBL-2H3 cells grown attached to the tissue culture flasks or as suspension cultures expressed the same integrin subunits identified in bone marrow, mesenteric and peritoneal mast cells ex vivo. Similarly, the expression of alpha4 integrin was higher in adherent cells. Therefore, alpha4 integrins may play a critical role in the anchorage of mast cells to the extracellular matrix in bone marrow and in peripheral tissues.

Characterization of ß-trypsin at acid pH by differential scanning calorimetry

Trypsin is a serino-protease with a polypeptide chain of 223 amino acid residues and contains six disulfide bridges. It is a globular protein with a predominance of antiparallel ß-sheet and helix in its secondary structure and has two domains with similar structures. We assessed the stability of ß-trypsin in the acid pH range using microcalorimetric (differential scanning calorimetry) techniques. Protein concentrations varied in the range of 0.05 to 2.30 mg/ml. Buffer solutions of 50.0 mM ß-alanine and 20.0 mM CaCl2 at different pH values (from 2.0 to 4.2) and concentrations of sorbitol (1.0 and 2.0 M), urea (0.5 M) or guanidinium hydrochloride (0.5 and 1.0 M) were used. The data suggest that we are studying the same conformational transition of the protein in all experimental situations using pH, sorbitol, urea and guanidinium hydrochloride as perturbing agents. The observed van't Hoff ratios (deltaHcal/deltaHvH) of 1.0 to 0.5 in the pH range of 3.2 to 4.2 suggest protein aggregation. In contrast, deltaHcal/deltaHvH ratios equal to one in the pH range of 2.0 to 3.2 suggest that the protein unfolds as a monomer. At pH 3.00, ß-trypsin unfolded with Tm = 54ºC and deltaH = 101.8 kcal/mol, and the change in heat capacity between the native and unfolded forms of the protein (deltaCp) was estimated to be 2.50 ± 0.07 kcal mol-1 K-1. The stability of ß-trypsin calculated at 298 K was deltaG D = 5.7 kcal/mol at pH 3.00 and deltaG D = 15.2 kcal/mol at pH 7.00, values in the range expected for a small globular protein.

Differential role of entorhinal and hippocampal nerve growth factor in short- and long-term memory modulation

We studied the effects of infusion of nerve growth factor (NGF) into the hippocampus and entorhinal cortex of male Wistar rats (250-300 g, N = 11-13 per group) on inhibitory avoidance retention. In order to evaluate the modulation of entorhinal and hippocampal NGF in short- and long-term memory, animals were implanted with cannulae in the CA1 area of the dorsal hippocampus or entorhinal cortex and trained in one-trial step-down inhibitory avoidance (foot shock, 0.4 mA). Retention tests were carried out 1.5 h or 24 h after training to measure short- and long-term memory, respectively. Immediately after training, rats received 5 µl NGF (0.05, 0.5 or 5.0 ng) or saline per side into the CA1 area and entorhinal cortex. The correct position of the cannulae was confirmed by histological analysis. The highest dose of NGF (5.0 ng) into the hippocampus blocked short-term memory (P < 0.05), whereas the doses of 0.5 (P < 0.05) and 5.0 ng (P < 0.01) NGF enhanced long-term memory. NGF administration into the entorhinal cortex improved long-term memory at the dose of 5.0 ng (P < 0.05) and did not alter short-term memory. Taken as a whole, our results suggest a differential modulation by entorhinal and hippocampal NGF of short- and long-term memory.

Differential effect of plant lectins on mast cells of different origins

Histamine release induced by plant lectins was studied with emphasis on the carbohydrate specificity, external calcium requirement, metal binding sites, and mast cell heterogeneity and on the importance of antibodies bound to the mast cell membrane to the lectin effect. Peritoneal mast cells were obtained by direct lavage of the rat peritoneal cavity and guinea pig intestine and hamster cheek pouch mast cells were obtained by dispersion with collagenase type IA. Histamine release was induced with concanavalin A (Con A), lectins from Canavalia brasiliensis, mannose-specific Cymbosema roseum, Maackia amurensis, Parkia platycephala, Triticum vulgaris (WGA), and demetallized Con A and C. brasiliensis, using 1-300 µg/ml lectin concentrations applied to Wistar rat peritoneal mast cells, peaking on 26.9, 21.0, 29.1, 24.9, 17.2, 10.7, 19.9, and 41.5%, respectively. This effect was inhibited in the absence of extracellular calcium. The lectins were also active on hamster cheek pouch mast cells (except demetallized Con A) and on Rowett nude rat (animal free of immunoglobulins) peritoneal mast cells (except for mannose-specific C. roseum, P. platycephala and WGA). No effect was observed in guinea pig intestine mast cells. Glucose-saturated Con A and C. brasiliensis also released histamine from Wistar rat peritoneal mast cells. These results suggest that histamine release induced by lectins is influenced by the heterogeneity of mast cells and depends on extracellular calcium. The results also suggest that this histamine release might occur by alternative mechanisms, because the usual mechanism of lectins is related to their binding properties to metals from which depend the binding to sugars, which would be their sites to bind to immunoglobulins. In the present study, we show that the histamine release by lectins was also induced by demetallized lectins and by sugar-saturated lectins (which would avoid their binding to other sugars). Additionally, the lectins also released histamine from Rowett nude mast cells that are free of immunoglobulins.

Differential activity of a lectin from Solieria filiformis against human pathogenic bacteria

A lectin isolated from the red alga Solieria filiformis was evaluated for its effect on the growth of 8 gram-negative and 3 gram-positive bacteria cultivated in liquid medium (three independent experiments/bacterium). The lectin (500 µg/mL) stimulated the growth of the gram-positive species Bacillus cereus and inhibited the growth of the gram-negative species Serratia marcescens, Salmonella typhi, Klebsiella pneumoniae, Enterobacter aerogenes, Proteus sp, and Pseudomonas aeruginosa at 1000 µg/mL but the lectin (10-1000 µg/mL) had no effect on the growth of the gram-positive bacteria Staphylococcus aureus and B. subtilis, or on the gram-negative bacteria Escherichia coli and Salmonella typhimurium. The purified lectin significantly reduced the cell density of gram-negative bacteria, although no changes in growth phases (log, exponential and of decline) were observed. It is possible that the interaction of S. filiformis lectin with the cell surface receptors of gram-negative bacteria promotes alterations in the flow of nutrients, which would explain the bacteriostatic effect. Growth stimulation of the gram-positive bacterium B. cereus was more marked in the presence of the lectin at a concentration of 1000 µg/mL. The stimulation of the growth of B. cereus was not observed when the lectin was previously incubated with mannan (125 µg/mL), its hapten. Thus, we suggest the involvement of the binding site of the lectin in this effect. The present study reports the first data on the inhibition and stimulation of pathogenic bacterial cells by marine alga lectins.

Differential evolution based classification with pool of distances and aggregation operators

The objective of this thesis is to develop and generalize further the differential evolution based data classification method. For many years, evolutionary algorithms have been successfully applied to many classification tasks. Evolution algorithms are population based, stochastic search algorithms that mimic natural selection and genetics. Differential evolution is an evolutionary algorithm that has gained popularity because of its simplicity and good observed performance. In this thesis a differential evolution classifier with pool of distances is proposed, demonstrated and initially evaluated. The differential evolution classifier is a nearest prototype vector based classifier that applies a global optimization algorithm, differential evolution, to determine the optimal values for all free parameters of the classifier model during the training phase of the classifier. The differential evolution classifier applies the individually optimized distance measure for each new data set to be classified is generalized to cover a pool of distances. Instead of optimizing a single distance measure for the given data set, the selection of the optimal distance measure from a predefined pool of alternative measures is attempted systematically and automatically. Furthermore, instead of only selecting the optimal distance measure from a set of alternatives, an attempt is made to optimize the values of the possible control parameters related with the selected distance measure. Specifically, a pool of alternative distance measures is first created and then the differential evolution algorithm is applied to select the optimal distance measure that yields the highest classification accuracy with the current data. After determining the optimal distance measures for the given data set together with their optimal parameters, all determined distance measures are aggregated to form a single total distance measure. The total distance measure is applied to the final classification decisions. The actual classification process is still based on the nearest prototype vector principle; a sample belongs to the class represented by the nearest prototype vector when measured with the optimized total distance measure. During the training process the differential evolution algorithm determines the optimal class vectors, selects optimal distance metrics, and determines the optimal values for the free parameters of each selected distance measure. The results obtained with the above method confirm that the choice of distance measure is one of the most crucial factors for obtaining higher classification accuracy. The results also demonstrate that it is possible to build a classifier that is able to select the optimal distance measure for the given data set automatically and systematically. After finding optimal distance measures together with optimal parameters from the particular distance measure results are then aggregated to form a total distance, which will be used to form the deviation between the class vectors and samples and thus classify the samples. This thesis also discusses two types of aggregation operators, namely, ordered weighted averaging (OWA) based multi-distances and generalized ordered weighted averaging (GOWA). These aggregation operators were applied in this work to the aggregation of the normalized distance values. The results demonstrate that a proper combination of aggregation operator and weight generation scheme play an important role in obtaining good classification accuracy. The main outcomes of the work are the six new generalized versions of previous method called differential evolution classifier. All these DE classifier demonstrated good results in the classification tasks.

A phonological and tonal analysis of Samue using Optimality Theory

The topic of the present doctoral dissertation is the analysis of the phonological and tonal structures of a previously largely undescribed language, namely Samue. It is a Gur language belonging to the Niger-Congo language phulym, which is spoken in Burkina Faso. The data were collected during the fieldwork period in a Sama village; the data include 1800 lexical items, thousands of elicited sentences and 30 oral texts. The data were first transcribed phonetically and then the phonological and tonal analyses were conducted. The results show that the phonological system of Samue with the phoneme inventory and phonological processes has the same characteristics as other related Gur languages, although some particularities were found, such as the voicing and lenition of stop consonants in medial positions. Tonal analysis revealed three level tones, which have both lexical and grammatical functions. A particularity of the tonal system is the regressive Mid tone spreading in the verb phrase. The theoretical framework used in the study is Optimality theory. Optimality theory is rarely used in the analysis of an entire language system, and thus an objective was to see whether the theory was applicable to this type of work. Within the tonal analysis especially, some language specific constraints had to be created, although the basic Optimality Theory principle is the universal nature of the constraints. These constraints define the well-formedness of the language structures and they are differently ranked in different languages. This study gives new insights about typological phenomena in Gur languages. It is also a fundamental starting point for the Samue language in relation to the establishment of an orthography. From the theoretical point of view, the study proves that Optimality theory is largely applicable in the analysis of an entire sound system.

Differential response to gepirone but not to chlordiazepoxide in malnourished rats subjected to learned helplessness

The learned helplessness (LH) paradigm is characterized by learning deficits resulting from inescapable events. The aims of the present study were to determine if protein-calorie malnutrition (PCM) alters learning deficits induced by LH and if the neurochemical changes induced by malnutrition alter the reactivity to treatment with GABA-ergic and serotonergic drugs during LH. Well-nourished (W) and PCM Wistar rats (61 days old) were exposed or not to inescapable shocks (IS) and treated with gepirone (GEP, 0.0-7.5 mg/kg, intraperitoneally, N = 128) or chlordiazepoxide (0.0-7.5 mg/kg, intraperitoneally, N = 128) 72 h later, 30 min before the test session (30 trials of escape learning). The results showed that rats exposed to IS had higher escape latency than non-exposed rats (12.6 ± 2.2 vs 4.4 ± 0.8 s) and that malnutrition increased learning impairment produced by LH. GEP increased the escape latency of W animals exposed or non-exposed to IS, but did not affect the response of PCM animals, while chlordiazepoxide reduced the escape deficit of both W and PCM rats. The data suggest that PCM animals were more sensitive to the impairment produced by LH and that PCM led to neurochemical changes in the serotonergic system, resulting in hyporeactivity to the anxiogenic effects of GEP in the LH paradigm.

The performance of semi-quantitative differential PCR is similar to that of real-time PCR for the detection of the MYCN gene in neuroblastomas

Amplification of the MYCN gene in neuroblastomas is a potent biological marker of highly aggressive tumors, which are invariably fatal unless sound clinical management is applied. To determine the usefulness of semi-quantitative differential PCR (SQ-PCR) for accurate quantification of MYCN gene copy number, we evaluated the analytical performance of this method by comparing the results obtained with it for 101 tumor samples of neuroblastoma to that obtained by absolute and relative real-time PCR. Similar results were obtained for 100 (99%) samples, no significant difference was detected between the median log10 MYCN copy number (1.53 by SQ-PCR versus 1.55 by absolute real-time PCR), and the results of the two assays correlated closely (r = 0.8, Pearson correlation; P < 0.001). In the comparison of SQ-PCR and relative real-time PCR, SQ-PCR versus relative real-time PCR concordant results were found in 100 (99%) samples, no significant difference was found in median log10 MYCN copy number (1.53 by SQ-PCR versus 1.27 by relative real-time PCR), and the results of the two assays correlated closely (r = 0.8, Pearson correlation; P < 0.001). These findings indicate that the performance of SQ-PCR was comparable to that of real-time PCR for the amplification and quantification of MYCN copy number. Thus, SQ-PCR can be reliably used as an alternative assay in laboratories without facilities for real-time PCR.

Differential gene expression profiles of hepatocellular carcinomas associated or not with viral infection

Chronic hepatitis B (HBV) and C (HCV) virus infections are the most important factors associated with hepatocellular carcinoma (HCC), but tumor prognosis remains poor due to the lack of diagnostic biomarkers. In order to identify novel diagnostic markers and therapeutic targets, the gene expression profile associated with viral and non-viral HCC was assessed in 9 tumor samples by oligo-microarrays. The differentially expressed genes were examined using a z-score and KEGG pathway for the search of ontological biological processes. We selected a non-redundant set of 15 genes with the lowest P value for clustering samples into three groups using the non-supervised algorithm k-means. Fisher’s linear discriminant analysis was then applied in an exhaustive search of trios of genes that could be used to build classifiers for class distinction. Different transcriptional levels of genes were identified in HCC of different etiologies and from different HCC samples. When comparing HBV-HCC vs HCV-HCC, HBV-HCC/HCV-HCC vs non-viral (NV)-HCC, HBC-HCC vs NV-HCC, and HCV-HCC vs NV-HCC of the 58 non-redundant differentially expressed genes, only 6 genes (IKBKβ, CREBBP, WNT10B, PRDX6, ITGAV, and IFNAR1) were found to be associated with hepatic carcinogenesis. By combining trios, classifiers could be generated, which correctly classified 100% of the samples. This expression profiling may provide a useful tool for research into the pathophysiology of HCC. A detailed understanding of how these distinct genes are involved in molecular pathways is of fundamental importance to the development of effective HCC chemoprevention and treatment.