Algorithms for Exact Structure Discovery in Bayesian Networks

Bayesian networks are compact, flexible, and interpretable representations of a joint distribution. When the network structure is unknown but there are observational data at hand, one can try to learn the network structure. This is called structure discovery. This thesis contributes to two areas of structure discovery in Bayesian networks: space--time tradeoffs and learning ancestor relations. The fastest exact algorithms for structure discovery in Bayesian networks are based on dynamic programming and use excessive amounts of space. Motivated by the space usage, several schemes for trading space against time are presented. These schemes are presented in a general setting for a class of computational problems called permutation problems; structure discovery in Bayesian networks is seen as a challenging variant of the permutation problems. The main contribution in the area of the space--time tradeoffs is the partial order approach, in which the standard dynamic programming algorithm is extended to run over partial orders. In particular, a certain family of partial orders called parallel bucket orders is considered. A partial order scheme that provably yields an optimal space--time tradeoff within parallel bucket orders is presented. Also practical issues concerning parallel bucket orders are discussed. Learning ancestor relations, that is, directed paths between nodes, is motivated by the need for robust summaries of the network structures when there are unobserved nodes at work. Ancestor relations are nonmodular features and hence learning them is more difficult than modular features. A dynamic programming algorithm is presented for computing posterior probabilities of ancestor relations exactly. Empirical tests suggest that ancestor relations can be learned from observational data almost as accurately as arcs even in the presence of unobserved nodes.

Mediapolulla - Kasvatustieteellistä ja teknisen alan osaamista yhdistämässä : Opiskelijoiden yhteistyö ja oppiminen tieto- ja viestintätekniikkaprojektissa

The study concentrated on interdisciplinary teamwork of students in Helsinki University Department of Education and Helsinki University of Technology. Students worked in small interdisciplinary groups (n 12) to plan and teach in an information- and communication technology (ICT) club in elementary schools. The focus of the study was co-operation in the student groups and students learning experiences. Theoretical background of the study consists of theories of collaboration and socially shared cognition. Study was an qualitative case study and the data was collected with individual focus interviews and learning diaries. The data was categorised and the connections between categories were analysed with a table. Shared cognition appeared as a form of distribution of tasks and in the actual processes of shared expertise. The tasks were shared according to students expertise. Processes of shared expertise were joint knowledge building, integration of interests, awareness and exploitation of others expertise and allowing freedom for others to use their expertise. Additionally expression of ones own views and setting an example to others were one sided sharing of expertise. Students of technology were responsible of technical issues and the responsibility sphere of educational science students was more fragmented. For instance they concentrated in taking children s abilities into consideration. The sphere of shared cognition included also the need for tutoring and learning from others. Usually students did not directly learn from representative of other discipline, instead the learning for instance of social skills happened indirectly. Learning was fostered if learning was set as a goal and prevented if the differences in expertise were too minor. Sharing of cognition was prevented if co-operation was too problematic. Co-operation was usually successful. Good planning, good person chemistry and appreciation of expertise of others promoted success. Problems caused by different backgrounds were usually slight. Successful interaction was complementary and equal. Groups were usually able to circumvent problems in communication and use of justification in discussion promoted co-operation. When comparing the groups in the scope of the study, two were found to be notably opposed and the other groups located between these extreme cases, but the elements of success prevailed. Learning experiences concentrated on social skills, project management, school world and ICT. Essential was achieved field experience and observation of ones own capabilities. In organisation of student interdisciplinary co-operation is important to ensure sufficient differences in expertise and guide students to gain complementary interaction and appropriate setting of goals. Interdisciplinary field project prepared students to face the demands of

Distribution and adhesion-promoting activity of alpha4 and alpha5 chain laminins in human endothelia and carcinomas

Basement membranes are specialized sheets of extracellular matrix found in contact with epithelia, endothelia, and certain isolated cells. They support tissue architecture and regulate cell behaviour. Laminins are among the main constituents of basement membranes. Due to differences between laminin isoforms, laminins confer structural and functional diversity to basement membranes. The first aim of this study was to gain insights into the potential functions of the then least characterized laminins, alpha4 chain laminins, by evaluating their distribution in human tissues. We thus created a monoclonal antibody specific for laminin alpha4 chain. By immunohistochemistry, alpha4 chain laminins were primarily localized to basement membranes of blood vessel endothelia, skeletal, heart, and smooth muscle cells, nerves, and adipocytes. In addition, alpha4 chain laminins were found in the region of certain epithelial basement membranes in the epidermis, salivary gland, pancreas, esophagus, stomach, intestine, and kidney. Because of the consistent presence of alpha4 chain laminins in endothelial basement membranes of blood vessels, we evaluated the potential roles of endothelial laminins in blood vessels, lymphatic vessels, and carcinomas. Human endothelial cells produced alpha4 and alpha5 chain laminins. In quantitative and morphological adhesion assays, human endothelial cells barely adhered to alpha4 chain-containing laminin-411. The weak interaction of endothelial cells with laminin-411 appeared to be mediated by alpha6beta1 integrin. The alpha5 chain-containing laminin-511 promoted endothelial cell adhesion better than laminin-411, but it did not promote the formation of cell-extracellular matrix adhesion complexes. The adhesion of endothelial cells to laminin-511 appeared to be mediated by Lutheran glycoprotein together with beta1 and alphavbeta3 integrins. The results suggest that these laminins may induce a migratory phenotype in endothelial cells. In lymphatic capillaries, endothelial basement membranes showed immunoreactivity for laminin alpha4, beta1, beta2, and gamma1 chains, type IV and XVIII collagens, and nidogen-1. Considering the assumed inability of alpha4 chain laminins to polymerize and to promote basement membrane assembly, the findings may in part explain the incomplete basement membrane formation in these vessels. Lymphatic capillaries of ovarian carcinomas showed immunoreactivity also for laminin alpha5 chain and its receptor Lutheran glycoprotein, emphasizing a difference between normal and ovarian carcinoma lymphatic capillaries. In renal cell carcinomas, immunoreactivity for laminin alpha4 chain was found in stroma and basement membranes of blood vessels. In most tumours, immunoreactivity for laminin alpha4 chain was also observed in the basement membrane region of tumour cell islets. Renal carcinoma cells produced alpha4 chain laminins. Laminin-411 did not promote adhesion of renal carcinoma cells, but inhibited their adhesion to fibronectin. Renal carcinoma cells migrated more on laminin-411 than on fibronectin. The results suggest that alpha4 chain laminins have a counteradhesive function, and may thus have a role in detachment and invasion of renal carcinoma cells.

Young Men s Sexual Behaviour in Finland and Estonia : Opportunities for prevention of sexually transmitted infections

Background: The incidence of sexually transmitted infections (STIs) in most EU states has gradually increased and the rate of newly diagnosed HIV cases has doubled since 1999. STIs differ in their clinical features, prognosis and transmission dynamics, though they do share a common factor in their mode of transmission −that is, human behaviour. The evolvement of STI epidemiology involves a joint action of biological, epidemiological and societal factors. Of the more immediate factors, besides timely diagnosis and appropriate treatment, STI incidence is influenced by population patterns of sexual risk behaviour, particularly the number of sexual partners and the frequency of unprotected intercourse. Assessment of sexual behaviour, its sociodemographic determinants and time-trends are important in understanding the distribution and dynamic of STI epidemiology. Additionally, in the light of the basic structural determinants, such as increased level of migration, changes in gender dynamics and impacts from globalization, with its increasing alignment of values and beliefs, can reveal future challenges related to STI epidemiology. STI case surveillance together with surveillance on sexual behaviour can guide the identification of preventive strategies, assess their effectiveness and predict emerging trends. The objective of this study was to provide base line data on sexual risk behaviour, self-reported STIs and their patterns by sociodemographic factors as well as associations of sexual risk behaviour with substance use among young men in Finland and Estonia. In Finland national population based data on adult men s sexual behaviour is limited. The findings are discussed in the context of STI epidemiology as well as their possible implications for public health policies and prevention strategies. Materials and Methods: Data from three different cross-sectional population-based surveys conducted in Finland and Estonia, during 1998 2005, were used. Sexual behaviour- and health-related questions were incorporated in two surveys in Finland; the Health 2000, a large scale general health survey, focussed on young adults, and the Military health behavioural survey on military conscripts participating in the mandatory military training. Through research collaboration with Estonia, similar questions to the Finnish surveys were introduced to the second Estonian HIV/AIDS survey, which was targeted at young adults. All surveys applied mail-returned, anonymous, self-administered questionnaires with multiple choice formatted answers. Results: In Finland, differences in sexual behaviour between young men and women were minor. An age-stratified analysis revealed that the sex-related difference observed in the youngest age group (18 19 years) levelled off in the age group 20 24 and almost disappeared among those aged 25 29. Marital status was the most important sociodemographic correlate for sexual behaviour for both sexes, singles reporting higher numbers of lifetime-partners and condom use. This effect was stronger for women than for men. However, of those who had sex with casual partners, 15% were married or co-habiting, with no difference between male and female respondents. According to the Military health behavioural survey, young men s sexual risk behaviour in Finland did not markedly change over a period of time between 1998 and 2005. Approximately 30−40% of young men had had multiple sex partners (more than five) in their lifetime, over 20% reported having had multiple sex partners (at least three) over the past year and 50% did not use a condom in their last sexual intercourse. Some 10% of men reported accumulation of risk factors, i.e. having had both, multiple sex partners and not used a condom in their last intercourse, over the past year of the survey. When differences and similarities were viewed within Finland and Estonia, a clear sociodemographic patterning of sexual risk behaviour and self-reported STIs was found in Finland, but a somewhat less consistent trend in Estonia. Generally, both, alcohol and drug use were strong correlates for sexual risk behaviour and self-reported STIs in Finland and Estonia, having a greater effect on engagement with multiple sex partners rather than unprotected intercourse or self-reported STIs. In Finland alcohol use, relative to drug use, was a stronger predictor of sexual risk behaviour and self-reported STIs, while in Estonia drug use predicted sexual risk behaviour and self-reported STIs stronger than alcohol use. Conclusions: The study results point to the importance for prevention of sexual risk behaviour, particularly strategies that integrate sexual risk with alcohol and drug use risks. The results point to the need to focus further research on sexual behaviour and STIs among young people; on tracking trends among general population as well as applying in-depth research to identify and learn from vulnerable and high-risk population groups for STIs who are exposed to a combination of risk factors.

Campylobacter Infections of Domestic Origin in Finland

Campylobacter jejuni and C. coli are the leading causes of human bacterial gastroenteritis in developed countries. Most human Campylobacter infections are sporadic and a seasonal peak in the distribution of infections can be seen in the summer months in several countries, including Finland. A variety of risk factors for Campylobacter infections have been identified; handling and eating poultry, drinking unpasteurized milk, contact with domestic animals, and travelling abroad. However, the relative importance of the different risk factors in sporadic cases of Campylobacter infection remains unknown. In most cases, the infection is self-limiting and no specific treatment is required. Campylobacter enteritis can cause a wide range of complications, including reactive arthritis (ReA) that is reported in 1-5% of the cases. Seven clinical microbiology laboratories serving different geographical areas of Finland, participated in this multi-centre study, conducted during a seasonal peak in 2002. In a matched case-control study, domestically-acquired sporadic Campylobacter infections from three geographical areas were collected. The final study comprised 100 cases and 137 controls. Risk factors for sporadic domestically-acquired Campylobacter infections were identified on the basis of a questionnaire; swimming in natural waters was found to be a novel risk factor for Campylobacter infection. Other independent risk factors were tasting or eating raw or undercooked meat and drinking untreated water from a dug well. The role of bacterial strain and host characteristics are not fully understood in Campylobacter infections. Exposure factors, demographical characteristics, and the serotype of the Campylobacter isolate may affect the severity of the enteritis. This cross-sectional study comprised 114 patients with C. jejuni enteritis, diagnosed in three clinical microbiology laboratories; most of the patients had participated in the previous case-control study. Swimming was associated with age ≤ 5 years and serotype Pen 6,7 was found significantly more often among patients reporting swimming. The geographical distribution among serotypes varied; serotype Pen 4-complex appeared more often in patients from urban areas and serotype Pen 21 among patients from more rural areas. Thus, risk factors and sources of infection for C. jejuni infection may vary among individuals depending on age and geographical location. The in vitro susceptibilities of C. jejuni and C. coli strains isolated from patients infected abroad (85 strains) or domestically (393 strains) revealed that susceptibility to erythromycin is still high, even among isolates of foreign origin. However, the novel antimicrobial agent telithromycin did not offer any advantage over erythromycin; isolates with high minimal inhibitory concentrations (MICs) for erythromycin also showed reduced susceptibility to telithromycin. Reduced susceptibility to fluoroquinolones was detected almost exclusively among isolates of foreign origin and half of these isolates with high MICs for fluoroquinolones also showed elevated MICs for doxycycline. Questionnaires concerning complications associated with C. jejuni enteritis were sent to patients two months after becoming ill; 201 patients from seven different geographical areas were included in the study. Musculoskeletal complications after C. jejuni infection were commonly reported by patients (39%). The incidence of classical ReA was 4% and that of Achilles enthesopathy and/or heel pain 9%. Other C. jejuni-associated reactive joint symptoms were commonly reported, however, due to their milder nature seldom seen and diagnosed by a physician. The severity of the enteritis may predict further complications; stomach ache during enteritis was connected to the development of later joint pain. Early antimicrobial treatment, within two days from the start of symptoms, shortened the duration of diarrhoea by two days but did not prevent later musculoskeletal complications. Campylobacter is an important human enteropathogen and causes a significant burden of illness. As the incidence of Campylobacter infections is high, the importance of the infection and the occurrence of complications will increase. This stresses the importance of understanding the risk factors for acquiring Campylobacter infection and how bacterial strain and host characteristics may affect the risk for infection. The role of antimicrobial treatment for acute Campylobacter enteritis seems to be marginal and should be used restrictively.

On probability-based inference under data missing by design

Whether a statistician wants to complement a probability model for observed data with a prior distribution and carry out fully probabilistic inference, or base the inference only on the likelihood function, may be a fundamental question in theory, but in practice it may well be of less importance if the likelihood contains much more information than the prior. Maximum likelihood inference can be justified as a Gaussian approximation at the posterior mode, using flat priors. However, in situations where parametric assumptions in standard statistical models would be too rigid, more flexible model formulation, combined with fully probabilistic inference, can be achieved using hierarchical Bayesian parametrization. This work includes five articles, all of which apply probability modeling under various problems involving incomplete observation. Three of the papers apply maximum likelihood estimation and two of them hierarchical Bayesian modeling. Because maximum likelihood may be presented as a special case of Bayesian inference, but not the other way round, in the introductory part of this work we present a framework for probability-based inference using only Bayesian concepts. We also re-derive some results presented in the original articles using the toolbox equipped herein, to show that they are also justifiable under this more general framework. Here the assumption of exchangeability and de Finetti's representation theorem are applied repeatedly for justifying the use of standard parametric probability models with conditionally independent likelihood contributions. It is argued that this same reasoning can be applied also under sampling from a finite population. The main emphasis here is in probability-based inference under incomplete observation due to study design. This is illustrated using a generic two-phase cohort sampling design as an example. The alternative approaches presented for analysis of such a design are full likelihood, which utilizes all observed information, and conditional likelihood, which is restricted to a completely observed set, conditioning on the rule that generated that set. Conditional likelihood inference is also applied for a joint analysis of prevalence and incidence data, a situation subject to both left censoring and left truncation. Other topics covered are model uncertainty and causal inference using posterior predictive distributions. We formulate a non-parametric monotonic regression model for one or more covariates and a Bayesian estimation procedure, and apply the model in the context of optimal sequential treatment regimes, demonstrating that inference based on posterior predictive distributions is feasible also in this case.