Modelling transcriptional regulation with Gaussian processes

Davies, V. and Husmeier, D. (2014) Modelling transcriptional regulation with Gaussian processes. In: Valente, A. X.C.N., Sarkar, A. and Gao, Y. (eds.) Recent Advances in Systems Biology Research. Nova Science Publishers: New York, pp. 157-184. ISBN 9781629487366

Abstract

A challenging problem in systems biology is the quantitative modelling of transcriptional regulation. Transcription factors (TFs), which are the key proteins at the centre of the regulatory processes, may be subject to post-translational modification, rendering them unobservable at the mRNA level, or they may be controlled outside of the subsystem being modelled. In both cases, a mechanistic model description of the regula- tory system needs to be able to deal with latent activity profiles of the key regulators. A promising approach to deal with these difficulties is based on using Gaussian processes to define a prior distribution over the latent TF activity profiles. Inference is based on the principles of non-parametric Bayesian statistics, consistently inferring the posterior distribution of the unknown TF activities from the observed expression levels of potential target genes. The present work provides explicit solutions to the differ- ential equations needed to model the data in this manner, as well as the derivatives needed for effective optimisation. The work further explores identifiability issues not fully shown in previous work and looks at how this can cause difficulties with inference. We subsequently look at how the method works on two different TFs, including looking at how the model works with a more biologically realistic mechanistic model. Finally we analyse the effect of more biologically realistic non-Gaussian noise on the biologically realistic model showing how this can cause a reduction in the accuracy of the inference.

• Item Type: Book Sections
• Status: Published
• Glasgow Author(s) Enlighten ID: Husmeier, Professor Dirk and Davies, Dr Vinny
• Authors: Davies, V., and Husmeier, D.
• College/School: College of Science and Engineering > School of Mathematics and Statistics > Statistics
• Publisher: Nova Science Publishers
• ISBN: 9781629487366
• Copyright Holders: Copyright © 2014 Nova Science Publishers
• Publisher Policy: Reproduced with the permission of the publisher