Expression of aquaporin isoforms during human and mouse tooth development.

Previously, we described the development of hyaluronan (HA) deposition in human tooth germ tissues that are consistent with water transport in different stages of tooth development. The aquaporins (AQP) constitute a family of membrane water channels that are expressed in many organs. However, there are no data available about the expression pattern of aquaporin water channels in dental structures. In the present study we have characterised the expression of six different aquaporin isoforms (AQP1-5, AQP-9) in developing human and mouse tooth germs by immunohistochemistry using isoform specific antibodies. In the "bell stage" AQP1 was expressed in endothelial cells of small vessels whereas no other structures of the tooth primordial were labeled. AQP2, AQP3 and AQP9 immunoreactivity was not observed in tooth germs, whereas strong AQP4 and AQP5 expression was observed in dental lamina, inner enamel epithelium, stratum intermedium, stellate reticulum and the outer enamel epithelium. Oral epithelium also exhibited AQP4 and AQP5 immunolabeling. During development of the matrices of the dental hard tissues AQP4 and AQP5 immunostaining was observed in the odontoblasts and their processes, as well as in the secretory ameloblast and their apical processes. Immunolabeling controls were negative. In conclusion, AQP4 and AQP5 are expressed in tooth germ tissues in early development in cells that previously have been shown to express HA and/or CD44, indicating that AQP water channels may play a role for ECM hydration during tooth development.

On classification capability of neural networks: a case study with otoneurological data.

We investigated the capability of multilayer perceptron neural networks and Kohonen neural networks to recognize difficult otoneurological diseases from each other. We found that they are efficient methods, but the distribution of a learning set should be rather uniform. Also it is important that the number of learning cases is sufficient. If the two mentioned conditions are satisfied, these neural networks are similarly efficient as some other machine learning methods. The conditions are known in the theory of neural networks [1,2], but not often taken seriously in practice. Both networks functioned as well, excluding the case with several input variables, where the Kohonen neural networks surpassed the perceptron.

Postural control as assessed with virtual reality.

We studied the use of virtual reality technology as a stimulus in balance examinations. A pilot study was made using a small group of healthy subjects to investigate the effect of alcohol and virtual reality stimulus on the subjects' balance. The tests showed that blood alcohol concentration accounted for almost 50% of the increased lateral body sway velocity. The new stimulus technique based on virtual reality technology seems to be effective and flexible for postural investigations.

Classification of patients on the basis of otoneurological data by using Kohonen networks.

Machine learning methods such as neural networks, decision trees and genetic algorithms can be useful to aid in the classification of patients. We tested Kohonen artificial neural networks, which are known to be effective for classification tasks. Our sample included patients with six different diseases. The Kohonen network algorithm recognized the four largest groups reliably, but the two smallest groups were too small for the method. Neural networks seem to be promising for the computer-aided classification of otoneurological patients provided that the number of patients used is sufficiently large.

Signaling and subcellular localization of the TNF receptor Edar.

Tabby and downless mutant mice have identical phenotypes characterized by deficient development of several ectodermally derived organs such as teeth, hair, and sweat glands. Edar, encoded by the mouse downless gene and defective in human dominant and recessive forms of autosomal hypohidrotic ectodermal dysplasia (EDA) syndrome, is a new member of the tumor necrosis factor (TNF) receptor superfamily. The ligand of Edar is ectodysplasin, a TNF-like molecule mutated in the X-linked form of EDA and in the spontaneous mouse mutant Tabby. We have analyzed the response of Edar signaling in transfected cells and show that it activates nuclear factor-kappaB (NF-kappaB) in a dose-dependent manner. When Edar was expressed at low levels, the NF-kappaB response was enhanced by coexpression of ectodysplasin. The activation of NF-kappaB was greatly reduced in cells expressing mutant forms of Edar associated with the downless phenotype. Overexpression of Edar did not activate SAPK/JNK nor p38 kinase. Even though Edar harbors a death domain its overexpression did not induce apoptosis in any of the four cell lines analyzed, nor was there any difference in apoptosis in developing teeth of wild-type and Tabby mice. Additionally, we show that the subcellular localization of dominant negative alleles of downless is dramatically different from that of recessive or wild-type alleles. This together with differences in NF-kappaB responses suggests an explanation for the different mode of inheritance of the different downless alleles.

Analysis of the imputed female urinary incontinence data for the evaluation of expert system parameters.

We evaluated parameters for an expert system which will be designed to aid the differential diagnosis of female urinary incontinence by using knowledge discovered from data. To allow the statistical analysis, we applied means, regression and Expectation-Maximization (EM) imputation methods to fill in missing values. In addition, complete-case analysis was performed. Logistic regression results from the imputed data were reasonable. The significant parameters were mostly those that are important in the diagnostic work-up. Moreover, directions of relations between the parameters and the stress, mixed and sensory urge diagnoses were as expected. Analysis with the complete reduced data set gave clearly insufficient results. Imputed values had a moderate agreement, but odds ratios and classification accuracies of logistic regression equations were similar. Results suggest that with these data, simpler methods may be used to allow multivariate analysis and knowledge discovery, when better methods, such as EM imputation, are unavailable. Cluster analysis detected clusters corresponding to the small normal class, but was unable to clearly separate the larger incontinence classes.

A novel machine learning program applied to discover otological diagnoses.

A novel machine learning system, Galactica, has been developed for knowledge discovery from databases. This system was applied to discover diagnostic rules from a patient database containing 564 cases with vestibular schwannoma, bening paroxysmal positional vertigo, Ménière's disease, sudden deafness, traumatic vertigo and vestibular neuritis diagnoses. The rules were evaluated using an independent testing set. The accuracy of rules for these diagnoses were 91%, 96%, 81%, 95%, 92% and 98%, respectively. Besides being accurate, the rules contained the five most important diagnostic questions identified in the earlier research. The knowledge presented with rules can be easily comprehended and verified.

Experiences of otoneurological expert system for vertigo.

We have developed an OtoNeurological Expert system (ONE) to aid the diagnostics of vertigo, to assist teaching and to implement the database for research. The database contains detailed information on the patient history, signs and test results necessary for the diagnostic work with vertiginous patients. The pattern recognition method was used in the reasoning process. Questions regarding symptoms, signs and test results are weighted and scored for each disease, and the most likely disease is recognized from the defined disease profiles. Uncertainties in reasoning, caused by missing information, were solved with a method resembling fuzzy logic. We have also applied adaptive computer applications, such as genetic algorithms and decision trees, in the reasoning process. In the validation the expert system ONE proved to be a sound decision maker, by solving 65% of the cases correctly, while the physicians' mean was 69%. To improve the expert system ONE further, a follow-up should be implemented for the patients, to ease the diagnostic work of some difficult diseases. The six diseases were detected with high accuracy also with adaptive learning methods and discriminant analysis. An expert system is a practical tool in otoneurology. We aim to construct a hybrid program for the reasoning, where the best reasoning method for each disease is used.

Application of artificial intelligence in audiology.

In this paper, machine learning methods based on artificial intelligence theory are applied to the computer-aided decision making of some otoneurological diseases, for example Ménière's disease. Three methods explored are decision trees, genetic algorithms and neural networks. By using such a machine learning method, the decision-making program is trained with a representative training set of cases and tested with another set. The machine learning methods are useful also for our otoneurological expert system, One, which is based on a pattern recognition approach. The methods are able to differentiate most of the cases tested between the six diseases included, provided that a sufficiently large training set is available.

TNF signaling via the ligand-receptor pair ectodysplasin and edar controls the function of epithelial signaling centers and is regulated by Wnt and activin during tooth organogenesis.

Ectodermal dysplasia syndromes affect the development of several organs, including hair, teeth, and glands. The recent cloning of two genes responsible for these syndromes has led to the identification of a novel TNF family ligand, ectodysplasin, and TNF receptor, edar. This has indicated a developmental regulatory role for TNFs for the first time. Our in situ hybridization analysis of the expression of ectodysplasin (encoded by the Tabby gene) and edar (encoded by the downless gene) during mouse tooth morphogenesis showed that they are expressed in complementary patterns exclusively in ectodermal tissue layer. Edar was expressed reiteratively in signaling centers regulating key steps in morphogenesis. The analysis of the effects of eight signaling molecules in the TGFbeta, FGF, Hh, Wnt, and EGF families in tooth explant cultures revealed that the expression of edar was induced by activinbetaA, whereas Wnt6 induced ectodysplasin expression. Moreover, ectodysplasin expression was downregulated in branchial arch epithelium and in tooth germs of Lef1 mutant mice, suggesting that signaling by ectodysplasin is regulated by LEF-1-mediated Wnt signals. The analysis of the signaling centers in tooth germs of Tabby mice (ectodysplasin null mutants) indicated that in the absence of ectodysplasin the signaling centers were small. However, no downstream targets of ectodysplasin signaling were identified among several genes expressed in the signaling centers. We conclude that ectodysplasin functions as a planar signal between ectodermal compartments and regulates the function, but not the induction, of epithelial signaling centers. This TNF signaling is tightly associated with epithelial-mesenchymal interactions and with other signaling pathways regulating organogenesis. We suggest that activin signaling from mesenchyme induces the expression of the TNF receptor edar in the epithelial signaling centers, thus making them responsive to Wnt-induced ectodysplasin from the nearby ectoderm. This is the first demonstration of integration of the Wnt, activin, and TNF signaling pathways.

Associations of FGF-3 and FGF-10 with signaling networks regulating tooth morphogenesis.

The morphogenesis and cell differentiation in developing teeth is governed by interactions between the oral epithelium and neural crest-derived ectomesenchyme. The fibroblast growth factors FGF-4, -8, and -9 have been implicated as epithelial signals regulating mesenchymal gene expression and cell proliferation during tooth initiation and later during epithelial folding morphogenesis and the establishment of tooth shape. To further evaluate the roles of FGFs in tooth development, we analyzed the roles of FGF-3, FGF-7, and FGF-10 in developing mouse teeth. In situ hybridization analysis showed developmentally regulated expression during tooth formation for Fgf-3 and Fgf-10 that was mainly restricted to the dental papilla mesenchymal cells. Fgf-7 transcripts were restricted to the developing bone surrounding the developing tooth germ. Fgf-10 expression was observed in the presumptive dental epithelium and mesenchyme during tooth initiation, whereas Fgf-3 expression appeared in the dental mesenchyme at the late bud stage. During the cap and bell stage, both Fgf-3 and Fgf-10 were intensely expressed in the dental papilla mesenchymal cells both in incisors and molars. It is of interest that Fgf-3 expression was also observed in the primary enamel knot, a putative signaling center of the tooth, whereas no transcripts were seen in the secondary enamel knots that appear in the tips of future cusps of the bell stage tooth germs. Down-regulation of Fgf-3 and Fgf-10 expression in postmitotic odontoblasts correlated with the terminal differentiation of the odontoblasts and the neighboring ameloblasts. In the incisors, mesenchymal cells of the cervical loop area showed partially overlapping expression patterns for all studied Fgfs. In vitro analyses showed that expression of Fgf-3 and Fgf-10 in the dental mesenchyme was dependent on dental epithelium and that epithelially expressed FGFs, FGF-4 and -8 induced Fgf-3 but not Fgf-10 expression in the isolated dental mesenchyme. Beads soaked in Shh, BMP-2, and TGF-beta 1 protein did not induce either Fgf-3 or Fgf-10 expression. Cells expressing Wnt-6 did not induce Fgf-10 expression. Furthermore, FGF-10 protein stimulated cell proliferation in the dental epithelium but not in the mesenchyme. These results suggest that FGF-3 and FGF-10 have redundant functions as mesenchymal signals regulating epithelial morphogenesis of the tooth and that their expressions appear to be differentially regulated. In addition, FGF-3 may participate in signaling functions of the primary enamel knot. The dynamic expression patterns of different Fgfs in dental epithelium and mesenchyme and their interactions suggest existence of regulatory signaling cascades between epithelial and mesenchymal FGFs during tooth development.

Usefulness of imputation for the analysis of incomplete otoneurologic data.

The usefulness of imputation in the treatment of missing values of an otoneurologic database for the discriminant analysis was evaluated on the basis of the agreement of imputed values and the analysis results. The data consisted of six patient groups with vertigo (N=564). There were 38 variables and 11% of the data was missing. Missing values were filled in with the means, regression and Expectation-Maximisation (EM) imputation methods and a random imputation method provided the baseline results. Means, regression and EM methods agreed on 41-42% of the imputed missing values. The level of agreement between these and the random method was 20-22%. Despite the moderate agreement between the means, regression and EM methods, the discriminant functions were similar and accurate (prediction accuracy 83-99%). The discriminant functions obtained from the randomly imputed data were also accurate having prediction accuracy 88-97%. Imputation seems to be a useful method for treating the missing data in this database. However, a lot of data was missing in otoneurologic tests, which are likely to be of less importance in the diagnosis of vertiginous patients. Consequently, the disagreement of the methods did not affect clearly the discriminant analysis, and, therefore, future research requires more complete data and advanced imputation methods.

Nearest neighbour classification with heterogeneous proximity functions.

Heterogeneous proximity functions are similarity or distance functions which process data differently according to the scale of attributes. We compared two Minkowskian distance functions with three heterogeneous proximity functions to test whether these functions were better in data sets with attributes of mixed type. Significant differences in nearest neighbour classification accuracy were found in 11 of all the 21 data sets. City-block and Euclidean distance functions outperformed Gower's similarity function and Heterogeneous Euclidean-Overlap Metric, while Heterogeneous Value Difference Metric (HVDM) was better than the other functions. HVDM classified mixed data best, because it treats nominal attributes more carefully than the other functions.

Discovering diagnostic rules from a neurotologic database with genetic algorithms.

Data on patients with Meniere's disease, vestibular schwannoma, traumatic vertigo, sudden deafness, benign paroxysmal positional vertigo, or vestibular neuritis were retrieved from the database of otoneurologic expert system ONE for the development and testing of a genetic algorithm (GA). The accuracy of the diagnostic rules in solving the test cases was 81%, 91%, 92%, 95%, 96%, and 98% for the respective diseases. The best rules retrieved from the GA were described by a set of questions with the most likely answers. The most important questions concerned the duration of hearing loss and the occurrence of head injury. The validity and structure of the rules created with a GA can be analyzed in detail. For rare diseases, some other reasoning process can be used, for example, case-based reasoning.

Otoneurological expert system for vertigo.

We have developed an otoneurological expert system (ONE) to aid the diagnostics of vertigo, to assist teaching and to implement a database for research. The ONE database is set to harvest data on patient history, signs and test results necessary for diagnostic work with vertiginous patients. A method based on pattern recognition was used in the reasoning process. Questions about symptoms, signs and test results are weighted and scored for each disease and the most likely disease is recognized from defined disease profiles. Missing information and uncertainties are solved with a method resembling fuzzy logic. ONE was validated by comparing diagnoses assessed by physicians with those provided by the system. It proved to be a valid decision-maker by solving 65% of the cases correctly, while the physicians' mean was 69%. To improve ONE further, a follow-up should be implemented for the patients, since diagnosing sudden deafness and Meniere's disease during the first visit is often impossible. We aim to obtain new information on diseases involving vertigo by applying adaptive computer applications, such as genetic algorithms, to the reasoning process.

Comparison of genetic algorithms and other classification methods in the diagnosis of female urinary incontinence.

Galactica, a newly developed machine-learning system that utilizes a genetic algorithm for learning, was compared with discriminant analysis, logistic regression, k-means cluster analysis, a C4.5 decision-tree generator and a random bit climber hill-climbing algorithm. The methods were evaluated in the diagnosis of female urinary incontinence in terms of prediction accuracy of classifiers, on the basis of patient data. The best methods were discriminant analysis, logistic regression, C4.5 and Galactica. Practically no statistically significant differences existed between the prediction accuracy of these classification methods. We consider that machine-learning systems C4.5 and Galactica are preferable for automatic construction of medical decision aids, because they can cope with missing data values directly and can present a classifier in a comprehensible form. Galactica performed nearly as well as C4.5. The results are in agreement with the results of earlier research, indicating that genetic algorithms are a competitive method for constructing classifiers from medical data.

Treatment of missing values with imputation for the analysis of otologic data.

Usefulness of imputation in the treatment of missing values in an otologic database was studied. Missing values were filled in with means (ME), regression (LR) and Expectation-Maximization (EM) imputation methods. A random imputation method (RA) provided baseline results. ME, LR and EM methods agreed on 41-42% of the imputed missing values. The level of agreement between these and RA method was 20-22%. Despite the moderate agreement, discriminant functions were similar and accurate (prediction accuracy 83-99%) for each diagnosis. A lot of data were missing in otoneurotologic tests which have less weight in the diagnosis of vertiginous patients. Consequently, the disagreement of the methods did not affect discriminant analysis. Inputation seems to be a useful method to treat missing data in this database, but future research requires more complete data and advanced imputation methods.

Expression of the anhidrotic ectodermal dysplasia gene is reduced in skin cancer coinciding with reduced E-cadherin.

X-linked anhidrotic ectodermal dysplasia (EDA) is characterized by defects in the development of hair, teeth, and sweat glands. We have recently cloned the gene for EDA by positional cloning. The EDA gene encodes a transmembrane protein with a putative role in epithelial mesenchymal interactions. Since EDA could play a role in cell-cell or cell-matrix adhesion, acantholytic skin diseases and several types of non-invasive and invasive skin cancers were studied using in situ hybridization. Because of the observation that the promoter region of the EDA gene contains a binding site for LEF-1, which is involved in the signaling through E-cadherin/beta catenin complex, we compared the expression of EDA with immunolocalization for E-cadherin (E-CD). EDA expression during hair growth cycle, in benign adnexal tumors, and neuroectoderm-derived nevus cells was also examined. Our findings indicate that EDA expression is less abundant in malignant tumors, including basal and squamous cell carcinomas and melanoma, and in acantholytic keratinocytes compared to normal epidermis. The reduction in expression also coincides with diminished E-CD staining in all malignant cell types and in acantholytic cells. Our results suggest that EDA protein functions in the regulation of epithelial cell contacts and that it may be associated with the E-CD signaling pathway.

A genetic-based machine learning system to discover the diagnostic rules for female urinary incontinence.

A machine learning system named Galactica has been developed which uses a genetic algorithm to discover the rules for an expert system from databases. Galactica devised accurate diagnostic rules for female urinary incontinence from difficult heterogeneous data. The percentages of correctly classified stress, mixed and sensory urge incontinence testing cases were 89, 86 and 87%, respectively. However, these rules were rather general, consisting of 4-6 out of 13 conditions available in the data. Diagnostic rules for stress and mixed incontinence extracted from straightforward homogeneous data were highly accurate, classifying 100% of testing cases correctly as well as being specific, having from 10 to 11 conditions. More specific, but less accurate, rules were found from heterogeneous data with a biased fitness function. All of the rules were correct, i.e. every condition in the rules had the expected value specified by the expert. Although, Galactica achieved a slightly better classification than the discriminant analysis, it is argued that the genetic approach is better than the statistical one, due to symbolic rules being comprehensible, whereas understanding a complex mathematical model requires statistical expertise.