NNTox: Gene Ontology-Based Protein Toxicity Prediction Using Neural Network

Area of Science:

• Bioinformatics
• Synthetic Biology
• Computational Biology

Background:

• Synthetic biology advancements enable routine custom gene product synthesis.
• Protein sequence manipulation can unintentionally create toxic proteins, posing research risks.
• Current toxicity prediction methods lack broad applicability due to specificity.

Purpose of the Study:

• To develop a generalizable method for predicting protein toxicity.
• To identify the relationship between Gene Ontology (GO) terms and protein toxicity.
• To build accurate predictor models for protein toxicity and specific toxicity types.

Main Methods:

• Extended general protein function prediction methods for toxicity assessment.
• Developed a neural network model (NNTox) utilizing predicted GO terms.
• Created a multi-label model for predicting specific protein toxicity types.

Main Results:

• NNTox successfully predicts the likelihood of protein toxicity based on GO terms.
• The multi-label model accurately identifies specific types of protein toxicity.
• Established a link between GO terms and protein toxicity, enhancing predictive capabilities.

Conclusions:

• The developed models offer a robust approach to predicting protein toxicity before synthesis.
• NNTox and the multi-label model reduce risks associated with genetically engineered proteins.
• This work advances bioinformatics tools for safer synthetic biology research.