Carmel Kent

Biomedical data integration – capturing similarities while preserving disparities

One of the challenges of healthcare data processing, analysis and warehousing is the integration of data gathered from disparate and diverse data sources. Promoting the adoption of worldwide accepted information standards along with common terminologies and the use of technologies derived from semantic web representation, is a suitable path to achieve that. To that end, the HL7 V3 Reference Information Model (RIM) [1] has been used as the underlying information model coupled with the Web Ontology Language (OWL) [2] as the semantic data integration technology. In this paper we depict a biomedical data integration process and demonstrate how it was used for integrating various data sources, containing clinical, environmental and genomic data, within Hypergenes, a European Commission funded project exploring the Essential Hypertension [3] disease model.

How Interactive is a Semantic Network? Concept Maps and Discourse in Knowledge Communities

Computer-mediated learning needs to be social too. Interactivity is a central construct for collaborative knowledge construction in online communities. We present an operationalized framework for measuring interactivity in online discussions, based on our view of interactivity as a socio-constructivist process. We hypothesize that the traditional design for online discussion platforms, with linear, chronologically threaded forums and bulletin boards, would result in less interactive behavioral patterns. We propose a semantic network topology to online discussions, which in turn reflects a social constructivist process. To that end, we developed Ligilo, an online discussion platform. Here, each discussion contribution and content item is expressed as a node in a semantic network of posts. We describe a field study comparing interactivity using threaded-based discussion and Ligilos semantic, networked based discussion. Initial results indicate higher interactivity in content creation patterns, suggesting learning, motivation and sustainability for discussion and community.

Semantic warehousing of diverse biomedical information

One of the main challenges of data warehousing within biomedical information infrastructures is to enable semantic interoperability between its various stakeholders as well as other interested parties. Promoting the adoption of worldwide accepted information standards along with common controlled terminologies is the right path to achieve that. The HL7 v3 Reference Information Model (RIM) is used to derive consistent health information standards such as laboratory, clinical health record data, problem- and goal-oriented care, public health and clinical research. In this paper we describe a RIM-based warehouse which provides (1) the means for data integration gathered from disparate and diverse data sources, (2) a mixture of XML and relational schemas and (3) a uniform abstract access and query capabilities serving both healthcare and clinical research users. Through the use of constrained standards (templates),, we facilitate semantic interoperability which would be harder to achieve if we only used generic standards in use cases with unique requirements. Such semantic warehousing also lays the groundwork for harmonized representations of data, information and knowledge, and thus enables a single infrastructure to serve analysis tools, decision support applications, clinical data exchange, and point-of-care applications. In addition, we describe the implementation of that semantic warehousing within Hypergenes, a European Commission funded project focused on Essential Hypertension, to illustrate the unique concepts and capabilities of our warehouse.

On the complexity of sparse exon assembly

Gene structure prediction is one of the most important problems in computational molecular biology. A combinatorial approach to the problem, denoted Gene Prediction via Spliced Alignment, was introduced by Gelfand, Mironov and Pevzner [5]. The method works by finding a set of blocks in a source genomic sequence S whose concatenation (splicing) fits a target gene T belonging to a homologous species. Let S,T and the candidate exons be sequences of size O(n). The innovative algorithm described in [5] yields an O(n3) result for spliced alignment, regardless of filtration mode. In this paper we suggest a new algorithm which targets the case where filtering has been applied to the data, resulting in a set of O(n) candidate exon blocks. Our algorithm yields an

Interactivity in online discussions and learning outcomes

O(n^2 \sqrt{n})

On the complexity of sparse exon assembly.

solution for this case.