Scalable Feature Selection Applications for Genome-Wide Association Studies of Complex Diseases

Personalized medicine will revolutionize our capabilities to combat disease. Working toward this goal, a fundamental task is the deciphering of geneticvariants that are predictive of complex diseases. Modern studies, in the formof genome-wide association studies (GWAS) have afforded researchers with the opportunity to reveal new genotype-phenotype relationships through the extensive scanning of genetic variants. These studies typically contain over half a million genetic features for thousands of individuals. Examining this with methods other than univariate statistics is a challenging task requiring advanced algorithms that are scalable to the genome-wide level. In the future, next-generation sequencing studies (NGS) will contain an even larger number of common and rare variants. Machine learning-based feature selection algorithms have been shown to have the ability to effectively create predictive models for various genotype-phenotype relationships. This work explores the problem of selecting genetic variant subsets that are the most predictive of complex disease phenotypes through various feature selection methodologies, including filter, wrapper and embedded algorithms. The examined machine learning algorithms were demonstrated to not only be effective at predicting the disease phenotypes, but also doing so efficiently through the use of computational shortcuts. While much of the work was able to be run on high-end desktops, some work was further extended so that it could be implemented on parallel computers helping to assure that they will also scale to the NGS data sets. Further, these studies analyzed the relationships between various feature selection methods and demonstrated the need for careful testing when selecting an algorithm. It was shown that there is no universally optimal algorithm for variant selection in GWAS, but rather methodologies need to be selected based on the desired outcome, such as the number of features to be included in the prediction model. It was also demonstrated that without proper model validation, for example using nested cross-validation, the models can result in overly-optimistic prediction accuracies and decreased generalization ability. It is through the implementation and application of machine learning methods that one can extract predictive genotype–phenotype relationships and biological insights from genetic data sets.

Genetics of inherited small vessel diseases – in search of a novel small vessel disease and modifiers of the clinical course of CADASIL

The genetic and environmental risk factors of vascular cognitive impairment are still largely unknown. This thesis aimed to assess the genetic background of two clinically similar familial small vessel diseases (SVD), CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy) and Swedish hMID (hereditary multi-infarct dementia of Swedish type). In the first study, selected genetic modifiers of CADASIL were studied in a homogenous Finnish CADASIL population of 134 patients, all carrying the p.Arg133Cys mutation in NOTCH3. Apolipoprotein E (APOE) genotypes, angiotensinogen (AGT) p.Met268Thr polymorphism and eight NOTCH3 polymorphisms were studied, but no associations between any particular genetic variant and first-ever stroke or migraine were seen. In the second study, smoking, statin medication and physical activity were suggested to be the most profound environmental differences among the monozygotic twins with CADASIL. Swedish hMID was for long misdiagnosed as CADASIL. In the third study, the CADASIL diagnosis in the Swedish hMID family was ruled out on the basis of genetic, radiological and pathological findings, and Swedish hMID was suggested to represent a novel SVD. In the fourth study, the gene defect of Swedish hMID was then sought using whole exome sequencing paired with a linkage analysis. The strongest candidate for the pathogenic mutation was a 3’UTR variant in the COL4A1 gene, but further studies are needed to confirm its functionality. This study provided new information about the genetic background of two inherited SVDs. Profound knowledge about the pathogenic mutations causing familial SVD is also important for correct diagnosis and treatment options.

Clinical and genetic markers in disease progression – a study among subjects with obstructive lung diseases

Asthma, COPD, and asthma and COPD overlap syndrome (ACOS) are chronic pulmonary diseases with an obstructive component. In COPD, the obstruction is irreversible and the disease is progressive. The aim of the study was to define and analyze factors that affected disease progression and patients’ well-being, prognosis and mortality in Chronic Airway Disease (CAD) cohort. The main focus was on COPD and ACOS patients. Retrospective data from medical records was combined with genetic and prospective follow-up data. Smoking is the biggest risk factor for COPD and even after the diagnosis of the disease, smoking plays an important role in disease development and patient’s prognosis. Sixty percent of the COPD patients had succeeded in smoking cessation. Patients who had managed to quit smoking had lower mortality rates and less psychiatric diseases and alcohol abuse although they were older and had more cardiovascular diseases than patients who continued smoking. Genetic polymorphism rs1051730 in the nicotinic acethylcholine receptor gene (CHRNA3/5) associated with heavy smoking, cancer prevalence and mortality in two Finnish independent cohorts consisting of COPD patients and male smokers. Challenges in smoking cessation and higher mortality rates may be partly due to individual patient’s genetic composition. Approximately 50% of COPD patients are physically inactive and the proportion was higher among current smokers. Physically active and inactive patients didn’t differ from each other in regard to age, gender or comorbidities. Bronchial obstruction explained inactivity only in severe disease. Subjective sensation of dyspnea, however, had very strong association to inactivity and was also associated to low health related quality of life (HRQoL). ACOS patients had a significantly lower HRQoL than either the patients with asthma or with COPD even though they were younger than COPD patients, had better lung functions and smaller tobacco exposure.