A novel seven-tier framework for the classification of MEFV missense variants using adaptive and rigid classifiers

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Clinical categorization of MEFV gene variants often conflicts with in silico predictions.
• Accurate classification of MEFV variants is crucial for understanding familial Mediterranean fever and related conditions.

Purpose of the Study:

• To develop a robust, seven-tier classification system for MEFV missense variants of unknown significance (VUS).
• To establish a generalized pipeline for classifying variants in other genes.
• To improve the accuracy of variant classification and identify novel functional regions within the MEFV gene.

Main Methods:

• Extracted 12,017 human MEFV gene variants from Ensembl, identifying 6034 missense variants.
• Utilized 42 in silico tools and machine learning algorithms (bagging and boosting methods) for variant classification.
• Developed a novel classification and clustering methodology, including a grayscale interpretation system.

Main Results:

• The XGBoost model achieved the highest accuracy (0.9882), outperforming other machine learning methods.
• The proportion of known variants increased from 6.9% to 29.4% (a 4.3-fold improvement) after applying the new methodology.
• Identified two novel hotspot regions and one tolerant site within the pyrin protein, providing functional insights.

Conclusions:

• The developed framework offers an innovative and accurate approach for MEFV VUS classification.
• This method enhances the identification of pathogenic variants and functional regions, aiding in precision genomic medicine.
• The pipeline is scalable for classifying variants in other genes, with potential broad applications in genetic diagnostics.