Prediction of Human Papillomavirus-Host Oncoprotein Interactions Using Deep Learning

Area of Science:

• Computational biology
• Virology
• Bioinformatics

Background:

• Human papillomavirus (HPV) drives disease via complex viral-host protein interactions, particularly involving the PI3K signaling pathway.
• Proteins such as AKT, IQGAP1, and MMP16 are implicated in the development of HPV-associated cancers.
• Traditional methods for studying protein-protein interactions (PPIs) are resource-intensive; computational approaches offer greater efficiency.

Purpose of the Study:

• To develop and validate a deep learning model for predicting protein-protein interactions (PPIs) between human papillomavirus (HPV) and host proteins.
• To identify specific interactions between HPV oncoproteins (E6 and E7) and host proteins involved in cancer development.

Main Methods:

• A Recurrent Neural Network (RNN) algorithm was trained using available HPV-host protein interaction data.
• The model was developed on the SPYDER platform utilizing Python libraries including TensorFlow, Scikit-learn, Pandas, and NumPy.
• Data was partitioned into training, validation, and testing sets at a 7:1:2 ratio.

Main Results:

• The deep learning model demonstrated strong performance, achieving an MCC score of 0.7937 and over 88% accuracy on other metrics.
• Predictions indicated interactions between HPV 31/18 E6 and E7 proteins with the host protein AKT.
• HPV31 E7 showed predicted interactions with IQGAP1 and MMP16, with high confidence scores.

Conclusions:

• The developed model effectively predicts interactions between HPV oncoproteins (E6, E7) and host proteins within the PI3K pathway.
• These predicted interactions suggest a role for viral proteins in AKT activation, potentially driving HPV-associated cancers.
• The model provides a robust tool for predicting interactomes, facilitating experimental validation and advancing understanding of HPV pathogenesis.