The PSINK Project
Recently, the German Federal Ministry for Education and Research (BMBF) has granted funding for the development of an automated information retrieval system for preclinical and clinical published data on spinal cord injury (SCI) treatments. A consortium of researchers in the fields of (a) spinal cord injury of the Center for Neuronal Regeneration (CNR) and the University Düsseldorf, (b) semantic computing of the CITEC excellence cluster at the Bielefeld University, (c) natural language processing at Stuttgart Universitiy and (d) applied statistics together with experts on the planning and monitoring of clinical trials of the Coordination Center for Clinical Trials of the University Clinics of Düsseldorf established the project PSINK (Preclinical Spinal Cord Injury Knowledge Base.
This project will develop and deploy an information system which provides actionable knowledge to neuroscientists and clinicians developing therapies for spinal cord injuries with the goal to translate experimental therapeutic concepts into clinical practice to support recovery of paraplegic patients. A central component in this information system will be a database containing information about potential therapies to treat Spinal Cord Injuries (SCI). Key characteristics and outcomes of these therapies are extracted from evidence in the vast existing literature. Clinical researchers and governmental admission authorities will benefit from this information system by being able to retrieve and assess SCI candidate therapies at large scale on the basis of the entire empirical evidence available in the peer reviewed biomedical literature. This database will include connections to existing established knowledge resources to ensure interoperability.
Translational neuroscience in the area of spinal cord injury, as for many other still incurable neurological diseases, faces a strong gap between the immense preclinical knowledge available on the one hand, and the lack of successful clinical trials in spinal cord injury therapy on the other. The main hypothesis explored in this proposal is that by mining the huge evidence available about preclinical experiments concerning SCI therapies from the the literature, the above mentioned gap can
be closed or at least diminished.
In the first three years of the project, an information system will be developed which enables researchers to gain thus far unmatched insights into the most important parameters affecting the prospective success of experimental therapies and to propose the most promising therapies, e.g., for acute and chronic paraplegia to be translated into the clinical context on the basis of an objective evidence-based scoring system.
In the fourth and fifth year of the project, i.e., the translation phase, the database will be made available to the public and used for preparing the design of a clinical study based on the insights delivered by the information system and scoring scheme developed in the project. We will evaluate in this phase in how far our system could have avoided previous unsuccessful clinical trials and identify the crucial parameters which lead to the failure.
The efficient construction of the database will be enabled by the application of semantic text technologies, turning unstructured information from large volumes of scientific publications into structured form. This includes an information extraction workflow to process textual information of a highly relational structure and a stack of analytical services operating on the resulting database. At the core of the information extraction process is an innovative interactive learning cycle geared towards lifelong improvement of the extraction model, while at the same time minimizing manual annotation and curation efforts. Both these goals are accommodated in a novel combination of distant supervision and active learning techniques in the context of undirected probabilistic graphical models. This will ensure the long-term sustainability of the system. In spite of being adapted to the use case of identifying relevant promising SCI therapies in neuroscience, the methodologies and workflows to be developed in the project will be generic and thus applicable to other domains as well.
Because this is a novel approach, we require the feedback of potential users. By completing this questionnaire you can provide a valuable contribution to the quality and usability of the knowledge base.