Autoimmune neurologic diseases

Oral

Deep Learning Predicts Gut Microbiota Associated with Untreated Multiple Sclerosis

Thursday, June 15
5:45 PM - 7:00 PM

The gut microbiota influences its host's immune system, both systemically and in the CNS, and influences blood brain barrier permeability and demyelination. Thus, it may be associated with Multiple Sclerosis (MS). Several studies have found differential abundances of bacterial taxa among relapse-remitting MS (RRMS) patients. However, some of the patients Have been treated and confound the analysis. Machine learning could be used to predict a microbiota associated with MS as well as screen for biomarkers for MS. However, the nature of microbiota sequencing studies makes this approach challenging; often the number of variables in each sample far exceeds the number of samples available, making the most powerful deep machine learning algorithms, support vector machines (SVM) and artificial neural networks, incompatible. Sparse SVM provides a solution to this problem by selecting only the most informative variables to use in its algorithms; thus, it may also be used to screen for bacterial biomarkers. A Sparse SVM model was used to; (1) predict who has MS, (2) find operational taxanomic units (OUTs) and subspecies bacterial markers for untreated RRMS patients. V4 regions of 16S rRNA reads of untreated MS patient gut microbiota samples were retrieved from two previous studies. A Sparse SVM model was trained with gut bacterial OTUs or subspecies abundance profiles as input variables. The Sparse SVM model screened for the most informative bacterial taxa whose abundance predicted patients who have RRMS with the highest accuracy, above 90%. Some of the positively selected subspecies are known to regulate systemic immunity. ÂÂ

Ali Mirza

University of Michigan Department of Immunology

Presentation(s):

    Send Email for Ali Mirza

    Qi Wu

    University of Michigan Department of Neurology

    Presentation(s):

      Send Email for Qi Wu

      Qin Wang

      University of Michigan Department of Neurology

      Presentation(s):

        Send Email for Qin Wang

        Catherine Dowling

        University of Michigan Department of Neurology

        Presentation(s):

          Send Email for Catherine Dowling

          Joel Begay

          University of Michigan School of Public Health - Department of Epidemiology

          Presentation(s):

            Send Email for Joel Begay

            Yang Mao-Draayer

            Associate Professor
            University of Michigan Department of Neurology

            Presentation(s):

              Send Email for Yang Mao-Draayer


              Assets

              Deep Learning Predicts Gut Microbiota Associated with Untreated Multiple Sclerosis



              Attendees who have favorited this

              Send Email for Deep Learning Predicts Gut Microbiota Associated with Untreated Multiple Sclerosis