Prediction of the human membrane proteome

Area of Science:

• Genomics
• Proteomics
• Computational Biology

Background:

• Membrane proteins are vital cellular components and significant pharmaceutical targets.
• Limited availability of experimentally determined membrane protein structures necessitates computational prediction methods.

Purpose of the Study:

• To computationally predict membrane proteins within the human genome using multiple topology prediction algorithms.
• To estimate the proportion of human genes encoding membrane proteins and characterize their transmembrane topology.

Main Methods:

• Analysis of 21,416 annotated human genes using seven distinct membrane protein topology prediction methods.
• Algorithms employed include hidden Markov models, neural networks, and support vector machines.
• A majority decision approach was used to consolidate predictions from multiple methods.

Main Results:

• Predicted gene counts for membrane proteins varied from 5508 to 7651 across different methods.
• An estimated 5539 human genes (approximately 26%) encode membrane proteins.
• The majority of predicted membrane proteins possess a single transmembrane region, with a notable subset having seven, including G-protein coupled receptors.

Conclusions:

• Computational prediction is essential for understanding membrane protein roles and for drug discovery.
• This study provides a comprehensive estimate of membrane proteins in the human proteome.
• A publicly accessible visualization tool is available for exploring predicted membrane protein topologies.