Neural Networks model biological evolution of faithful epigenetic inheritance

Abstract The layer of histone Post-Translational Modification (PTM) patterns, present above the DNA strand, forms an important epigenetic marker sequence which regulates gene expression. The specific pattern of histone PTMs in the region of chromatin housing the gene is critical for turning on/off the expression of the corresponding gene. During DNA replication in mitotic cells, the available evidence suggests that the histone PTMs from the mother chromatid are transferred uniformly at random among the two daughter chromatids. Parental epigenetic memory as well as interactions among multiple PTMs at the same histone facilitates the reconstruction of the PTM sequence at the daughter chromatids. We show that this biological marvel aided by the epigenetic memory has evolutionary analogs in the sense that it can be learnt by an appropriate extended neural network. We show through simulations that high fidelity reconstruction of the mother chromatin’s patterns for certain PTMs can be achieved by our network. This model can be enhanced to include several more interacting histone PTMs, elucidating the role of each. The proposed neural network can possibly be used in a multitude of biological applications related to gene expression regulation.