Machine Reasoning And Visualization Based On Example Of Glial-Synapse Interplay Throughout Inflammation Dynamics

There are several separate observations found in the research communication. Pro-inflammatory stimuli impact glial cells that response with TNFa, IL-6 and IL-1β, which inhibit glutamate uptake by glial cells [Huichun Tong et al. 2017]. High glutamate concentrations stimulate extrasynaptic NMDA receptors that lead to the calcium influx. [Zhang et al. 2016]. Consequently, elevated cytosolic calcium and activated calpain preceded neurodegeneration [Kurbatskaya et al. 2016]. But no single complete experimental paper describes the whole molecular-level processes of the neuroinflammation (as of beginning of 2018 and to our best knowledge).

We utilized machine learning methods to match complementing facts from different papers. The algorithm generates single reasoning chain to hypothesize description of the biological process with the highest possible level of detalization. The reasoning module is a part of the sci.AI platform and analyzes lexical and biological groups of features of the tuples that are semantically extracted from the original research papers. The inferred pathway is visually encoded with spatial (histological) and timeline dimensions. It is done with the aim of communicating the process as close as possible to biological ground truth.

Here we present performance of the machine method based on the example of inferred inflammatory cascade and glial-synapse interactions.

The Poster was presented at Federation of European Neuroscience Societies (FENS) Forum of Neuroscience, Berlin, Germany, 2018