Prediction of Nucleosome Positioning Using Machine Learning Methods

Nucleosome is the basic structure of chromatin in eukaryotic cells. The location of nucleosomes plays an important role in gene transcriptional regulation. Knowledge of the detailed organization of nucleosomes in the genome and the mechanisms of nucleosome positioning is crucial for understanding gene regulation and expression. In recent years, a large number of works in the field of bioinformatics, physics and biology have been devoted to the study and prediction of nucleosome positioning in the genomes of various organisms based on many factors (frequency of nucleotide k-mers, geometric and physic-chemical properties of dinucleotides, analysis of primary DNA sequences with convolution neural networks). This work is devoted to the study and prediction of the nucleosome positioning in the human genome based on the structural properties and parameters of the electrostatic potential and bendability of the DNA sequence using machine learning models. Using the constructed model the generalizing probabilistic profile of nucleosome positioning was constructed and comparison of predictions in different types of tissues was made.