Self-organizing tree-growing network for the classification of protein sequences

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning

Background:

• Phylogenetic tree construction is crucial for understanding evolutionary relationships.
• Existing methods may struggle with highly diversified sequence datasets.
• The self-organizing tree algorithm (SOTA) offers a novel approach based on Kohonen's self-organizing maps and Fritzke's growing cell structures.

Purpose of the Study:

• To adapt and evaluate the self-organizing tree algorithm (SOTA) for analyzing sequence patterns beyond pre-aligned data.
• To demonstrate SOTA's capability in constructing high-resolution phylogenetic trees.
• To showcase SOTA's effectiveness in classifying diverse biological sequence datasets.

Main Methods:

• The self-organizing tree algorithm (SOTA) was adapted to analyze sequence patterns based on residue frequency, including protein dipeptide and other n-gram compositions.
• The algorithm's node generation was controlled by a similarity threshold to achieve desired phylogenetic resolution.
• SOTA was applied to analyze protein families (cytochrome c, triosephophate isomerase, hemoglobin alpha chains) and a mixture of interleukins and their receptors.

Main Results:

• SOTA successfully constructed accurate phylogenetic trees for protein families.
• The algorithm demonstrated proficiency in classifying highly diversified sequence datasets, including mixtures of interleukins and receptors.
• The adapted SOTA effectively handles sequence pattern analysis based on n-gram compositions.

Conclusions:

• The adapted self-organizing tree algorithm (SOTA) is a powerful tool for phylogenetic analysis and classification of biological sequences.
• SOTA's flexibility in analyzing n-gram compositions enhances its utility for diverse and complex biological datasets.
• This approach provides a robust method for evolutionary and classification studies in bioinformatics.