DeepNup: Prediction of Nucleosome Positioning from DNA Sequences Using Deep Neural Network

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Nucleosome positioning is crucial for regulating DNA accessibility to proteins, impacting various cellular processes.
• DNA sequence intrinsic preferences significantly influence nucleosome positioning.
• Accurate computational prediction of nucleosome positioning solely from DNA sequences remains challenging.

Purpose of the Study:

• To develop and validate a deep learning-based computational method, DeepNup, for improved nucleosome positioning prediction using DNA sequence information.
• To verify the contribution of DNA sequences to nucleosome positioning.

Main Methods:

• A hybrid feature encoding scheme combining One-hot encoding and Trinucleotide composition encoding for DNA sequences.
• Multiscale convolutional neural network modules with parallel convolution kernels and gated recurrent units to learn local and global features.
• A fully connected layer and sigmoid unit for classification and final prediction of nucleosome positioning.

Main Results:

• DeepNup demonstrated superior performance in nucleosome positioning prediction compared to existing state-of-the-art methods on benchmark datasets.
• The method effectively learns high-order feature representations from DNA sequences.
• Experimental evaluation metrics confirmed the efficacy of DeepNup.

Conclusions:

• DeepNup is a powerful deep learning tool for accurate identification of potential nucleosome sequences based solely on DNA sequence data.
• The study highlights the significant role of DNA sequence information in nucleosome positioning.
• DeepNup offers a valuable approach for advancing research in genomics and gene regulation.