iGHBP: Computational identification of growth hormone binding proteins from sequences using extremely randomised tree

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Growth hormone binding protein (GHBP) modulates growth hormone signaling, crucial for cellular mechanisms.
• Accurate GHBP identification is vital for understanding cell growth processes.
• The postgenomic era necessitates automated computational methods for identifying GHBP in large protein datasets.

Purpose of the Study:

• To develop a novel, accurate, and automated computational predictor for identifying growth hormone binding proteins (GHBP).
• To enhance the efficiency of GHBP identification within extensive protein sequence databases.

Main Methods:

• Developed iGHBP, a machine-learning predictor utilizing an extremely randomized tree algorithm.
• Employed a two-step feature selection protocol combining dipeptide composition and amino acid index values.
• Optimized feature selection to improve prediction accuracy.

Main Results:

• The iGHBP predictor achieved 84.9% accuracy during cross-validation, outperforming a control predictor by ~7%.
• Feature selection protocol proved effective in enhancing prediction performance.
• iGHBP demonstrated superior performance on an independent dataset compared to existing methods.

Conclusions:

• The iGHBP predictor offers a fast and accurate method for identifying GHBP from protein sequences.
• The developed feature selection strategy is effective for improving computational prediction models.
• A user-friendly web server for iGHBP is available for broader research application.