SENSE: Siamese neural network for sequence embedding and alignment-free comparison.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Traditional sequence alignment methods struggle with large-scale biological data.
• Existing alignment-free methods offer computational efficiency but lack precision in sequence similarity assessment.
• Heuristic-based approaches provide only approximations of true alignment distances.

Purpose of the Study:

• To develop an efficient and accurate alignment-free sequence comparison method.
• To introduce a deep learning-based approach for sequence embedding and analysis.
• To overcome the limitations of existing alignment-free techniques in approximating sequence distances.

Main Methods:

• Developed SENSE (SiamEse Neural network for Sequence Embedding), a deep neural network model.
• Trained the model on a small dataset to learn an explicit embedding function.
• Projected sequences into an embedding space to minimize the mean square error between alignment and embedding distances.

Main Results:

• SENSE significantly outperforms state-of-the-art alignment-free methods.
• The method demonstrates superior efficiency and accuracy in sequence comparison.
• This represents the first application of deep learning to alignment-free sequence analysis.

Conclusions:

• SENSE provides a powerful new tool for large-scale sequence analysis.
• The deep learning approach enables more precise and efficient sequence similarity assessment.
• Open-source software for SENSE is available for broader research application.