Browsing by Title

Rapid growth and advancement of next generation sequencing (NGS) technologies have changed the landscape of genomic medicine. Today, clinical laboratories perform DNA sequencing on a regular basis, which is an error prone process. Erroneous data affects downstream analysis and produces fallacious result. Therefore, external quality assessment (EQA) of laboratories working with NGS data is crucial. Validation of variations such as single nucleotide polymor- phism (SNP) and InDels (<50 bp) is fairly accurate these days. However, detection and quality assessment of large changes such as the copy number variation (CNV) continues to be a concern. In this work, we aimed to study the feasibility of an automated CNV concordance analysis for the laboratory EQA services. We benchmarked variants reported by 25 laboratories against the highly curated gold standard for the son (HG002/NA24385) of the askenazim trio from the Personal Genome Project published by the Genome in a Bottle Consortium (GIAB). We employed two methods to conduct concordance of CNVs, the sequence based comparison with Truvari and the in-house exome-based comparison. For deletion calls of two whole genome sequencing (WGS) submissions, Truvari gained a value greater than 88% and 68% for precision and recall respectively. Conversely, the in-house method’s precision and recall score peaked at 39% and 7.9% respectively for one WGS submission for both deletion and duplication calls. The results indicate that automated CNV concordance analysis of the deletion calls for the WGS-based callset might be feasible with Truvari. On the other hand, results for panel-based targeted sequencing for the deletion calls showed precision and recall rates ranging from 0-80% and 0-5.6% respectively with Truvari. The result suggests that automated concordance analysis of CNVs for targeted sequencing remains a challenge. In conclusion, CNV concordance analysis depends on how the sequence data is generated.

Puumala orthohantavirus (PUUV) is a single stranded negative sense RNA virus, carried by the bank vole (Myodes glareolus). Like other orthohantaviruses, it does not cause visible symptoms in the host species, but when transmitted to humans, it can cause a mild version of hemorrhagic fever with renal syndrome (HFRS) called nephropathia epidemica (NE). PUUV is the only pathogenic orthohantavirus that is endemic to Finland, where it has a relatively high incidence of approximately 35 in 100 000 inhabitants or 1000 to 3000 diagnosed cases annually. Here we describe a miniaturized immunofluorescence assay (mini-IFA) for measuring antibody response against PUUV from bank vole whole blood and heart samples as well as from patient serum samples. The method outline was based on the work done by Pietiäinen et al., (2022), but it was adapted for the detection of PUUV antibodies. Transfected cells expressing the PUUV structural proteins (N, GPC, Gn and Gc) were used instead of PUUV infected cells, which allowed for performing all steps outside of bio-safety level 3 (BSL3) conditions. This method also enables the simultaneous measurement of IgM, IgA and IgG antibody response from each sample in a more efficient and higher output manner, when compared to traditional immunofluorescence methods. Our results show that the method is effective for testing large amounts of samples for PUUV antibodies and it allows for quick and convenient access to high-quality images that can be used for both detecting interesting targets for future studies, as well as producing a visual archive of the test results.

Real estate appraisal, or property valuation, requires strong expertise in order to be performed successfully, thus being a costly process to produce. However, with structured data on historical transactions, the use of machine learning (ML) enables automated, data-driven valuation which is instant, virtually costless and potentially more objective compared to traditional methods. Yet, fully ML-based appraisal is not widely used in real business applications, as the existing solutions are not sufficiently accurate and reliable. In this study, we introduce an interpretable ML model for real estate appraisal using hierarchical linear modelling (HLM). The model is learned and tested with an empirical dataset of apartment transactions in the Helsinki area, collected during the past decade. As a result, we introduce a model which has competitive predictive performance, while being simultaneously explainable and reliable. The main outcome of this study is the observation that hierarchical linear modelling is a very potential approach for automated real estate appraisal. The key advantage of HLM over alternative learning algorithms is its balance of performance and simplicity: this algorithm is complex enough to avoid underfitting but simple enough to be interpretable and easy to productize. Particularly, the ability of these models to output complete probability distributions quantifying the uncertainty of the estimates make them suitable for actual business use cases where high reliability is required.

Nowadays the Internet is being used as a platform for providing a wide variety of different services. That has created challenges related to scaling IT infrastructure management. Cloud computing is a popular solution for scaling infrastructure, either by building a self-hosted cloud or by using cloud platform provided by external organizations. This way some the challenges related to large scale can be transferred to the cloud administrators. OpenStack is a group of open-source software projects for running cloud platforms. It is currently the most commonly used software for building private clouds. Since initially published by NASA and Rackspace, it has been used by various organizations such as Walmart, China Mobile and Cern nuclear research institute. The largest production deployments of OpenStack clouds consist of thousands of physical server computers located in multiple datacenters. The OpenStack community has created many deployment methods that take advantage of automated software configuration management. The deployment methods are built with state of the art software for automating different administrative tasks. They take different approaches to automating infrastructure management for OpenStack. This thesis compares some of the automated deployment methods for OpenStack and examines the benefits of using automation for configuration management. We present comparisons based on technical documentations as well as reference literature. Additionally, we conducted a questionnaire for OpenStack administrators about the use of automation. Lastly, we tested one of the deployment methods in a virtualized environment.

The assessment of nonverbal interaction is currently based on observations, interviews and questionnaires. The quantitative methods for assessment of nonverbal interaction are few. Novel technology allows new ways to perform assessment, and new methods are constantly being developed. Many of them are based on movement tracking by sensors, cameras and computer vision. In this study the use of OpenPose, a pose estimation algorithm, was investigated in detection of nonverbal interactional events. The aim was to find out whether the same meaningful interactional events could be found from videos by the algorithm and by human annotators. Another purpose was to find out the best way to annotate the videos in a study like this. The research material consisted of four videos of a child and a parent blowing soap bubbles. The videos were first run by OpenPose to track the poses of the child and the parent frame by frame. The data obtained by the algorithm was further processed by Matlab to extract the activities of the child and the parent, the coupling of the activities and the closeness of child’s and parent’s hands at each time point. The videos were manually annotated in two different ways: Both the basic units, such as the gaze directions and thehandling soap bubble jar, and the interactional events, such as communication initiatives, turn-taking and joint attention, were annotated. The results obtained by the algorithm were visually compared to annotations. The communication initiatives and turn-taking could be seen as peaks in hand closeness and as alternation in activities. However, interaction events were not the only reasons that caused changes in hand closeness and in activities, so they could not be distinguished from other actions solely by these factors. There also existed interaction that was not related to jar handling, which could not be seen from the hand closeness curves. With current recording arrangements, the gaze directions could not be detected by the algorithm and therefore the moments of joint attention could not be determined either. In order to enable the detection of gaze directions in the future studies, the faces of subjects are visible all the time. Distinguishing individual interaction events may not be the best way to assess interaction, and the focus of assessment should be in global units, such as synchrony between interaction partners. The best way to annotate the videos depends on the aim of the study.

Hybrid Clouds are one of the most notable trends in the current cloud computing paradigm and bare-metal cloud computing is also gaining traction. This has created a demand for hybrid cloud management and abstraction tools. In this thesis I identify shortcomings in Cloudify’s ability to handle generic bare-metal nodes. Cloudify is an open- source vendor agnostic hybrid cloud tool which allows using generic consumer-grade computers as cloud computing resources. It is not however capable to automatically manage joining and parting hosts in the cluster network nor does it retrieve any hardware data from the hosts, making the cluster management arduous and manual. I have designed and implemented a system which automates cluster creation and management and retrieves useful hardware data from hosts. I also perform experiments using the system which validate its correctness, usefulness and expandability.

Tämä Pro gradu -tutkielma käsittelee monivalintamuotoisten aukkotehtävien automaattista generointia suomen kielen sanataivutuksen harjoittelua varten. Aukkotehtävät ovat suosittu formaatti kielen opiskelussa ja kielitaidon arvioinnissa. Koska ne ovat muodoltaan melko hyvin kontrolloituja, niiden laatimisen automatisointi on ollut useiden akateemisten ja kaupallisten projektien tavoitteena viimeisten parin vuosikymmenen ajan. Tehtävä on osoittautunut haasteelliseksi. Jos aukkotehtävä generoidaan yksinkertaisesti poistamalla lauseesta sana, ja antamalla sen täyttäminen opiskelijalle tehtäväksi, käy helposti niin, ettei tehtävä ole mielekäs: usein näin tuotettuun aukkoon sopii monta vaihtoehtoista sanaa tai rakennetta. Yksi suurimmista haasteista aukkotehtävien generoinnissa on siis niin sanottu “aukkojen luotettavuus”: sen varmistaminen, että aukkoon sopiva ja epäsopiva vastaus pystytään erottamaan toisistaan. Yksi tapa varmistaa tämä on rajoittaa mahdollisten vastausten joukkoa antamalla vastausvaihtoehtoja, joiden tiedetään olevan vääriä. Tällöin automaattisen generoinnin haasteeksi nousee vääräksi tiedettyjen vaihtoehtojen löytäminen. Väärät vaihtoehdot eivät kuitenkaan saa olla sitä liian ilmeisellä tavalla: oikean vaihtoehdon valitsemisen täytyy muodostaa mielekäs haaste opiskelijalle. Tutkielmani pääasiallinen tavoite on tutkia luotettavien ja potentiaalisesti haastavien monivalintamuotoisten aukkotehtävien generoimista suomen kielen sanataivutuksen opiskelua varten. Kokeellisessa osiossa testaamaani metodia on aiemmin sovellettu menestyksekkäästi verrattavaan tarkoitukseen englannin kielen prepositioiden kontekstissa. Metodissa etsitään suuresta tekstikorpuksesta sellaisia prepositioita, jotka esiintyvät usein yhden aukon kontekstisanan kollokaationa, mutteivat koskaan kahden kontekstisanan kollokaationa samaan aikaan. Tavoitteeni on osoittaa, että metodia voi soveltaa myös suomen kielen taivutustehtävien generoimiseen. Testaan myös erityyppisten korpusten käyttöä tehtävän suorittamisessa, nimittäin yhtäältä peräkkäisyyteen perustuvia n-grammeja ja toisaalta syntaktiseen dependenssirakenteeseen perustuvia n-grammeja. Kokeellisen työn lisäksi erittelen työssäni kattavasti erilaisia tapoja muodostaa taivutusaukkotehtäviä, ja esittelen keksimäni aukkotehtävämallin. Keskeisin löydökseni on, että kyseisellä metodilla pystyy lisäämään aukkotehtävien luotettavuutta merkittävästi: sellaisissa testitapauksissa, joissa käytetty data on muutaman yksinkertaisen kriteerin mukaisesti arvioituna riittävää, jopa 80 % alun perin epäluotettavista aukoista muuttuu luotettaviksi. Lopussa pohdin tehtävien haasteellisuuden evaluointia sekä riittämättömän datan kysymyksiä. Mitä jälkimmäiseen tulee, argumentoin, että vaikka esille tulleiden datan riittävyyteen liittyvien haasteiden ratkaiseminen parantaisi tuloksia nykyisestään, voi metodia pitää tarkoitukseen sopivana jo sellaisenaan.

Social media provides huge amounts of potential data for natural language processing but using this data may be challenging. Finnish social media text differs greatly from standard Finnish and models trained on standard data may not be able to adequately handle the differences. Text normalization is the process of processing non-standard language into its standardized form. It provides a way to both process non-standard data with standard natural language processing tools and to get more data for training new tools for different tasks. In this thesis I experiment with bidirectional recurrent neural network models and models based on the ByT5 foundation model, as well as the Murre normalizer to see if existing tools are suitable for normalizing Finnish social media text. I manually normalize a small set of data from the Ylilauta and Suomi24 corpora to use as a test set. For training the models I use the Samples of Spoken Finnish corpus and Wikipedia data with added synthetic noise. The results of this thesis show that there are no existing tools suitable for normalizing Finnish written on social media. There is a lack of suitable data for training models for this task. The ByT5-based models perform better than the BRNN models.

The importance of Automatic Speech Recognition cannot be underestimated in today’s worlds as they play a significant role in human computer interaction. ASR systems have been studied deeply over time, but their maximum potential is yet to be explored for the Finnish language. Development of a traditional ASR system involves a number of hand-crafted engineering which has made this technology quite difficult and resourceful to develop. However, with advancements in the field of neural networks, end-to-end ASR neural networks can be developed which can automatically learn the mappings of audio to its corresponding transcript., therefore reducing hand crafted engineering requirements. End-to-end neural network ASR systems have been largely developed commercially by tech giants such as Microsoft, Google and Amazon. However, there are limitations to these commercial services such as data privacy and cost of usage. In this thesis, we explored existing studies in the development of an end-to-end neural network ASR for Finnish language. One successful technique utilized in the development of neural network ASR in the advent of inadequate data is Transfer learning. This is the approach explored in this thesis for the development of the end-to-end neural network ASR system. In addition, the success of this approach was evaluated. In order to achieve this purpose, dataset collected from the Finnish Bank of Finland and Kaggle were used to fine-tune Mozilla DeepSpeech model which is a pretrained end-to-end neural network ASR in English language. The results obtained by fine-tuning the pretrained neural network ASR in English for Finnish language showed a word error rate as low as 40% and character error rate as low as 22%. We therefore concluded that transfer learning is a successful technique for creating ASR model for a new language using a pretrained model in another language with little effort, data and resources.

Digitalisaation myötä rahoituspalvelut tulevat yhä suuremman joukon saataville. Palvelujen digitaalinen käyttö lisääntyy jatkuvasti, ja markkinatoimijat innovoivat uusia liiketoimintamalleja. Digitaaliset palvelut lisäävät mittakaavaetuja, jolloin yhä laadukkaampia palveluja on saatavilla edullisemmin. EU:n tavoitteena on edistää digitaalisten palvelujen käyttöönottoa kuluttajien ja yritysten eduksi ja tulla johtajaksi digitaaliseksi toimijaksi. Samalla muutokseen liittyviä riskejä on pyrittävä hallinnoimaan sijoittajansuoja varmistaen. Digitalisaation myötä myös sijoitusneuvonta on alkanut automatisoitua. Automatisoidussa sijoitusneuvonnassa sijoitussuosituksen tuottaa algoritmi, eikä ihmisten välistä vuorovaikutusta ole välttämättä ollenkaan. Automatisaatio demokratisoi sijoituspalveluja ja siihen liittyy huomattavia tehokkuusetuja palvelun edullisuuden ansiosta. Toisaalta uudenlainen liiketoimintamalli saattaa tuoda mukanaan sijoittajansuojaa vaarantavia riskejä. Tutkielmassa on pyritty hahmottamaan automatisoitua sijoitusneuvontaa koskeva eurooppaoikeudellinen sääntelykehys, jolloin keskeiseen asemaan nousee rahoitusvälineiden markkinoista annettu direktiivi (MiFID II). Sama sääntely kattaa niin perinteisen kuin automatisoidunkin sijoitusneuvonnan, minkä vuoksi tutkielmassa on pyritty luomaan digitaalisten palveluiden erityispiirteet huomioon ottavia tulkintasuosituksia. Tutkielman ollessa eurooppaoikeudellinen, tulkinnassa korostuvat EU-oikeuden tavoitteet, periaatteet ja EU-oikeudelle tyypillinen tulkintametodiikka, erityisesti teleologinen tulkinta. EU-oikeuden tavoitteiden taloudellisen luonteen vuoksi tutkielmassa painottuu oikeustaloustiede sekä paikoin myös oikeustaloustieteellisesti orientoitunut sääntelyteoria. Sijoituspalvelun tarjoajaan kohdistuvien MiFID II -direktiivin mukaisten velvoitteiden lisäksi tutkielmassa tarkastellaan relevantin sääntelyn roolia kansallisessa täytäntöönpanossa. Automatisoitua sijoitusneuvontaa rasittaa palvelun rajat ylittävän luonteen vuoksi korostunut sääntelyarbitraasin riski, joka johtuu kansallisen täytäntöönpanon eriävyydestä jäsenvaltioiden välillä. Lisäksi oikeuksien täytäntöönpanoon liittyy tiettyjä teknologian käytöstä johtuvia ongelmia, kuten uudenlaisia vastuukysymyksiä ja algoritmin läpinäkymättömyydestä johtuvia ongelmia näytön esittämisessä. Tältä osin tutkielma esittelee MiFID II -direktiiville tulkintatapoja, jotka johtavat mahdollisimman yhdenmukaiseen oikeuksien täytäntöönpanoon. Tutkielmassa päädytään esittämään, että teleologisesti tulkiten sama sääntely voi onnistuneesti olla sovellettavissa niin perinteiseen kuin automatisoituun sijoitusneuvontaan. Tulkinta ei kaikissa tapauksissa ole ongelmatonta, mutta uutta EU-sääntelyä on aiheen tiimoilta jo tulossa. Tuleva sääntely saattaa vähentää niin tutkielmassa havaittuja tulkintaongelmia kuin sääntelyarbitraasin riskiä.