PRCFX-DT: a new graph-based approach for feature selection and classification of genomic sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Viral diseases pose significant global health threats, necessitating advanced analytical methods.
• Analyzing viral genomic sequences offers insights into viral evolution and characteristics.
• Graph algorithms and machine learning show promise in analyzing viral data.

Purpose of the Study:

• To develop a novel approach for viral genomic sequence analysis using complex networks and probabilistic graph modeling.
• To extract distinguishing features from viral nucleotide sequences for improved identification.
• To evaluate the efficacy of graph-based methods in differentiating viral types and variants.

Main Methods:

• Utilized complex networks and probabilistic graph modeling for feature extraction from viral genomic sequences.
• Applied the PageRank centrality algorithm on codons associated with nucleotide sequences.
• Employed machine learning algorithms, specifically a decision tree classifier, for sample differentiation.

Main Results:

• The graph-based approach successfully extracted distinguishing features from viral genomic sequences.
• A decision tree classifier achieved 99.73% accuracy in differentiating 30 virus types, including coronavirus variants.
• The extracted graph node centrality features demonstrated high discriminative capability and relevance to genetic concepts.

Conclusions:

• The proposed graph-based algorithm effectively analyzes viral genomic sequences for feature extraction and structural analysis.
• The method shows significant potential for identifying any virus type or specific viral variant.
• Interpretability of extracted features aids in understanding the structural basis of viral identification.