Mathematical Sciences Research Seminar Series

Machine learning approaches for genetic prediction of schizophrenia

open book on table
Seminars

As part of the School of Science and Technology Mathematical Sciences Research Seminar Series, Matthew Smith, University of Cardiff presents: Machine learning approaches for genetic prediction of schizophrenia.

  • From: Wednesday 21 November 2018, 1 pm
  • To: Wednesday 21 November 2018, 2 pm
  • Location: 169, New Hall Block, Nottingham Trent University, Clifton Campus, Clifton Lane, Nottingham, NG11 8NS

Past event

Event details

As part of the School of Science and Technology Mathematical Sciences Research Seminar Series, Matthew Smith, University of Cardiff presents: Machine learning approaches for genetic prediction of schizophrenia.

Abstract

Schizophrenia is a complex, highly heritable psychiatric disorder. Attempts to predict schizophrenia from genetic data typically rely on the use of polygenic risk scores (PRS), which combine the effects of weakly-associated single nucleotide polymorphisms (SNPs) into a single score by assuming a linear additive model. Supervised machine learning encompasses a range of methods for learning patterns from labelled data; their success at prediction in other fields has sparked interest in their application to medical data, especially psychiatry. Here we investigate the use of linear and radial basis function (RBF) kernel support vector machines (SVMs), random forests (RFs), gradient boosted machines (GBMs), neural networks and multivariate logistic regression for prediction of a binary outcome using simulations of independent or interacting SNPs. We show that while RBF-SVMS, RFs and GBMs can learn complex genetic interactions in low dimensions, they do not out-perform PRS on independent SNPs. Classifiers were further investigated for their ability to predict schizophrenia using genetic and non-genetic factors, and for the effects of class imbalance on predictions, using data from individuals in the UK Biobank. We observed poor performance from individual SNPs alone, but a much-improved prediction when combining genetics with environmental and demographic factors.

All Welcome

This seminar is hosted by Professor Nadia Chuzhanova

For any enquiries please contact Dr David Chappell.

Location details

Room/Building:

169, New Hall Block

Address:

Nottingham Trent University
Clifton Campus
Clifton Lane
Nottingham
NG11 8NS

Past event

Still need help?

+44 (0)115 941 8418