Inferring the perturbation time from biological time course data

• DOI: 10.1093/bioinformatics/btw329
• Publication Date: 2016-06-11T03:54:37Z
• Authors: Jing Yang; Christopher A. Pe...; Murray R. Grant; Magnus Rattray
• Keywords: 0301 basic medicine; 0303 health sciences; Models, Statistical; Models, Genetic; Transcription, Genetic; QK; Arabidopsis; Pseudomonas syringae; Bayes Theorem; Quantitative Biology - Quantitative Methods; Original Papers; 03 medical and health sciences; FOS: Biological sciences; SB; Algorithms; Quantitative Methods (q-bio.QM)

Abstract Motivation: Time course data are often used to study the changes a biological process after perturbation. Statistical methods have been developed determine whether such perturbation induces over time, e.g. comparing perturbed and unperturbed time dataset uncover differences. However, existing do not provide principled statistical approach identify specific when two datasets first begin diverge perturbation; we call this time. Estimation of for different variables in allows us sequence events following therefore provides valuable insights into likely causal relationships. Results: We propose Bayesian method infer given from wild-type system. use non-parametric based on Gaussian Process regression. derive probabilistic model noise-corrupted replicated coming same profile before diverging The likelihood function can be worked out exactly posterior distribution is obtained by simple histogram approach, without recourse complex approximate inference algorithms. validate simulated apply it transcriptional change occurring Arabidopsis inoculation with Pseudomonas syringae pv. tomato DC3000 versus disarmed strain DC3000hrpA. Availability Implementation: An R package, DEtime, implementing available at https://github.com/ManchesterBioinference/DEtime along code required reproduce all results. Contact: [e-mail] or [e-mail] Supplementary information: Bioinformatics online.