Identification of cyclin protein using gradient boost decision tree algorithm.

Area of Science:

• Biochemistry
• Molecular Biology
• Bioinformatics

Background:

• Cyclin proteins regulate the cell cycle by forming complexes with cyclin-dependent kinases.
• Accurate identification of cyclin proteins is essential for studying their functions.
• Existing sequence similarity methods struggle due to low sequence homology among cyclins.

Purpose of the Study:

• To develop a computational model for accurate identification of cyclin proteins.
• To discriminate cyclin proteins from non-cyclin proteins using machine learning.

Main Methods:

• Protein sequences were encoded using seven feature types, including amino acid composition and various correlation measures.
• Feature optimization was performed using analysis of variance (ANOVA) and minimum redundancy maximum relevance (mRMR) with incremental feature selection (IFS).
• A gradient boost decision tree (GBDT) classifier was trained on the selected optimal features.

Main Results:

• The developed model achieved a high accuracy of 93.06%.
• The model demonstrated a strong performance with an Area Under the Curve (AUC) value of 0.971.
• The model's performance surpassed that of two recent studies on the same dataset.

Conclusions:

• The proposed machine learning model effectively identifies cyclin proteins.
• This computational approach offers a significant improvement over traditional sequence similarity methods.
• The model provides a valuable tool for advancing research in cell cycle regulation.