Photosynthetic protein classification using genome neighborhood-based machine learning feature

Area of Science:

• Plant biology
• Computational biology
• Biochemistry

Background:

• Identifying photosynthetic proteins is crucial for enhancing photosynthetic efficiency.
• Genome neighborhood analysis offers valuable insights for protein function prediction.

Purpose of the Study:

• To develop a computational method for identifying photosynthetic proteins using machine learning and genome neighborhood features.
• To improve the accuracy and efficiency of photosynthetic protein classification.

Main Methods:

• Utilized genome neighborhood network (GNN) patterns for feature extraction.
• Applied machine learning algorithms, specifically Random Forest (RF), for protein classification.
• Evaluated performance using accuracy and Mathew's correlation coefficient (MCC).

Main Results:

• The RF classifier achieved 87% accuracy in photosynthetic protein classification.
• The GNN-based method demonstrated superior performance (MCC = 0.718) compared to sequence similarity search (0.447) and other ML methods (0.361).
• The model successfully identified novel photosynthetic proteins.

Conclusions:

• Genome neighborhood profiles can infer protein functions, aiding in photosynthetic protein identification.
• The developed GNN-based ML method is effective for classifying photosynthetic proteins and discovering new ones.
• The tool is publicly available for broader research applications.