Home // SIMUL 2020, The Twelfth International Conference on Advances in System Simulation // View article
Authors:
Wim de Mulder
Rafel Riudavets
Martin Kuiper
Keywords: Recurrent neural networks; Nash-Sutcliffe efficiency; Epigenetics.
Abstract:
Recurrent Neural Networks (RNN) have been used in multiple tasks such as speech recognition, music composition and protein homology detection. In particular, they have shown superior performance in predicting structure in time series data. To our knowledge, RNN have not been used on DNA methylation data. Methylation patterns on chromosomal DNA represent an important form of epigenetic imprinting, a form of epigenetics that results in heritable gene expression and phenotype changes. DNA methylation is one of the mechanisms that a cell uses to fine-tune the expression levels of its individual genes, and it has been shown to affect very specific areas around specific genes. The methylation state of the human chromosomal DNA can be readily assessed with microarray technology, allowing the determination of the methylation status of thousands of positions along the individual chromosomes of the genome. With RNN analysis we show that these methylation patterns have substantial structure, when relatively large stretches of chromosomes are tested. Furthermore, we show that each chromosome appears to have its own distinctive sequential methylation structure, but that this structure breaks down, to some extent, when normal cells develop into a tumour.
Pages: 48 to 53
Copyright: Copyright (c) IARIA, 2020
Publication date: October 18, 2020
Published in: conference
ISSN: 2308-4537
ISBN: 978-1-61208-831-0
Location: Porto, Portugal
Dates: from October 18, 2020 to October 22, 2020