A Supervised Learning Approach for Accurate and Efficient Identification of Chikungunya Virus Lineages and Signature...

Area of Science:

• Virology
• Genomics
• Machine Learning
• Epidemiology

Background:

• Chikungunya virus (CHIKV) is a significant public health concern.
• Continuous viral evolution requires high-resolution genomic surveillance.
• Existing methods struggle to efficiently differentiate CHIKV sub-lineages.

Purpose of the Study:

• To develop CHIKVGenotyper, an interpretable machine learning framework for high-resolution CHIKV lineage classification.
• To provide an accurate and scalable tool for CHIKV molecular epidemiology.
• To gain insights into viral evolution and support outbreak response.

Main Methods:

• Leveraged a dataset of 6886 CHIKV genome sequences, refining it to 3014 for model development.
• Developed a hierarchical assignment pipeline integrating sequence matching, machine learning, and phylogenetic validation.
• Trained and evaluated multiple machine learning models, including Random Forest, and utilized SHAP analysis.

Main Results:

• The optimal Random Forest model achieved near-perfect accuracy (F1-score: 99.53%) on whole-genome data and robust performance (F1-score: 96.50%) on low-coverage data.
• The E2 glycoprotein alone demonstrated comparable accuracy (F1-score: 99.52%), indicating its discriminative potential.
• Identified key lineage-defining mutations (e.g., E1-K211E, E2-V264A) using SHAP analysis, consistent with biological knowledge.

Conclusions:

• CHIKVGenotyper offers an accurate, scalable, and interpretable solution for high-resolution CHIKV lineage classification.
• The tool enhances molecular epidemiology by providing insights into viral evolution.
• Findings support improved outbreak response strategies through precise genomic surveillance.