ABSTRACT
On the 21st century, the exponential growth of technology, led the world facing a myriad of information coming from multitudinous sources. Then, finding ways of storing knowledge committed to certain rules became imperious. Ontologies have been playing an important role on connecting data to the semantics of the real world. Data, without such ontological commitment, could be interpreted as representations of different entities than the one it actually is, leading to biased analysis and inaccurate prediction on data-driven projects. Such kind of artifact formalizes shared knowledge regarding a domain of discourse. Therefore, this study will, based on works showing the benefits of bringing ontologies to the scenario of Machine Learning techniques, enrich similarity metrics between instances of data. So, the Human Disease Ontology (DO) will be used. Instead of calculating pairwise similarities between two diseases (terms on DO), groups of diseases will be considered. Therefore, this work will rely on adapting a groupwise similarity metric Data collection will be done considering the SIVEP-Gripe Dataset. Then, an analysis will be made on how better Machine Learning Algorithms can perform the analysis is made considering semantic rather than just numerical and categorical features. © 2022 Copyright for this paper by its authors.
ABSTRACT
The relevance of foundational ontologies and well-founded conceptual models is acknowledged in several contexts for making the real-world semantics of data explicit, and for reducing the impact of semantic ambiguities during the integration and manipulation of different data sources. Few domains pose such demands as urgently as the analysis and knowledge extraction from COVID-19 data. Since COVID-19 was declared a pandemic in early 2020, huge efforts from around the world provided an avalanche of data for research and analysis at an unprecedented rate. However, the coexistence of semantically divergent and non-explicit definitions for data from distinct countries and time periods that are being integrated and analyzed makes the conclusions of such analysis and the extracted knowledge potentially questionable. This work contributes to the development of a preliminary version of OntoCOVID, an ontology for the domain of COVID-19 well-founded in UFO and built using the SABiO methodology. © 2021 Copyright for this paper by its authors.
ABSTRACT
COVID-19 is an infectious disease caused by the newly discovered coronavirus named SARS-CoV-2. The virus enters the body through the airways by exploiting the angiotensin-converting enzyme 2 (ACE2) and serine proteinase TMPRSS2. Thus, especially lung epithelial cells are attacked by the virus. In the distal lung, the virus infection leads to life-threatening alveolar damage and cytokine storm. Many excellent clinical studies described the pathology of COVID-19 progression in patients. However, impactful in vitro studies are still missing due to the exceptional difficulty to model the alveolar setting in vitro. Here, we introduce two advanced models based on organoids and lung-on-chip (LOC) technology. The models derived from primary alveolar epithelial cells were characterized by fluorescence imaging and gene expression. Barrier function and cell polarity were assessed by confocal microscopy of tight junction and transporter proteins. We showed that the LOC model and the newly developed organoids preserve alveolar type II specific markers (SP-C and HTII-280). In both models we could detect relevant amounts of ACE2 and TMPRSS2 mRNA as compared to whole lung extracts. As a next step we will optimize the infection titer for both systems to further analyze transcriptomic changes upon SARS-CoV-2 infection. The here presented advanced in vitro models, recapitulating the distal lung, serve as complementary SARS-CoV-2 lung infection models. The LOC model will allow to simulate alveolar breakdown and cytokine storm whereas the organoid models will facilitate high-content drug screening.