Knowledge Mining

An essential activity in AETIONOMY is to generate hypotheses about multiscale mechanisms of neurodegenerative pathophysiology. Conceptually, we identified and organized disease specific features, at different scales, to perform data-driven analysis. This analysis serves to identify robust combinations of features that correspond to disease sub-types. The mechanisms of neurodegenerative pathophysiology, that distinguish the disease subtypes – referred to as our hypotheses - will be tested, iteratively elaborated and validated, through data generated by apposite patient studies.

Figure: Criteria of disease taxonomies are mainly based on the effects of the disease process rather than aetiological mechanisms, which may help to understand biological processes of the diseases

The AETIONOMY Knowledge base (AKB) provides a platform that:

  • presents all approaches to identify new mechanistic disease hypotheses from different sources with stories and webinar recordings, and

  • integrates all methods, disease models, web services (tools) and data resources including knowledge from literature (extracted by text mining).
    Our starting point for knowledge mining was to execute queries on our Information Retrieval and Text Mining system SCAIView. The goal of this processing is to extract information on diseases and their specific indicators (genes, SNPs, biomarkers, stages, etc.). So here text mining methods were applied, ranging from statistically validated co-occurrence detection and shallow parsing, the extraction of dependency relation expressed in the syntactic structure of sentences. The latter is based on the predicate-argument structures (triples), and can capture relationships between the entities even if they occur in substantial distance within the sentence.

Webinar: A webinar about the Knowledge Management Platform (D2.4.3.3) is in preparation. Afterwards a web-recording will be available here.

Figure: Knowledge Mining using SCAView and BELIEF Information Retrieval