Using Machine Learning for Predicting the Effect of Mutations in the Initiation Codon.

Area of Science:

• Genetics
• Bioinformatics
• Molecular Biology

Background:

• Traditional mutation effect prediction focuses on amino acid substitutions.
• This approach is insufficient for mutations in translation initiation codons, which can halt protein synthesis.
• Specific predictive methods are required for initiation codon mutations.

Purpose of the Study:

• To develop a novel method for predicting the effects of mutations in the canonical translation initiation codon.
• To create a predictor that utilizes a biological model incorporating mutation-specific features.

Main Methods:

• Developed a predictor using tree-based machine learning algorithms.
• Utilized data extracted from Ensembl for model training.
• Incorporated biological features, including distance to alternative initiation codons.

Main Results:

• The developed model accurately detects deleterious or benign mutations in the canonical translation initiation codon.
• Achieved a precision of 44.28% and an accuracy of 98.32%.
• Outperforms existing state-of-the-art tools like PolyPhen, SIFT, and CADD for initiation codon mutations.

Conclusions:

• The proposed method offers improved prediction of mutation effects at the translation initiation codon.
• This biological model-based approach addresses a critical gap in mutation effect prediction.
• The findings enhance the understanding and prediction of genetic variations impacting protein synthesis initiation.