Machine learning approaches demonstrate that protein structures carry information about their genetic coding.

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Synonymous codons, which encode the same amino acid, were traditionally considered functionally equivalent in protein synthesis.
• Emerging evidence suggests a link between synonymous codon usage and protein structure, challenging this long-held assumption.
• Understanding this relationship is crucial for fields like protein engineering and synthetic biology.

Purpose of the Study:

• To investigate the association between synonymous codon identity and protein backbone dihedral angles.
• To determine if codon sequences offer superior predictive power for protein structure compared to amino acid sequences.
• To explore local codon-codon dependencies and identify the most informative codon context for predicting synonymous codon identity.

Main Methods:

• Employed regression and classification models to analyze the relationship between codon sequences and protein structural features.
• Compared the predictive accuracy of codon sequences versus amino acid sequences for protein backbone dihedral angles.
• Utilized a classification approach to assess codon context, evaluating the predictive power of local codon-codon interactions.

Main Results:

• Codon sequences demonstrated a lower error rate in predicting protein backbone dihedral angles than amino acid sequences.
• Models trained with actual dihedral angles showed improved classification of synonymous codons compared to those trained with random angles.
• Identified specific codon context positions that significantly enhance the prediction accuracy of synonymous codon identity.

Conclusions:

• Synonymous codon identity is not inconsequential and carries significant information about protein structure.
• Codon sequence provides a more accurate basis for predicting protein backbone dihedral angles than amino acid sequence alone.
• Local codon context, particularly specific positions, plays a critical role in determining synonymous codon usage and its structural implications.