Near perfect protein multi-label classification with deep neural networks

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Biological sequences, such as proteins, are high-dimensional data.
• Classifying and comparing these sequences against large databases is crucial for biological research.
• Artificial neural networks (ANNs) are increasingly successful in various data processing tasks.

Purpose of the Study:

• To develop and evaluate novel ANNs for multi-label classification of biological sequences.
• To assess the accuracy of these ANNs in classifying protein sequences into established biological families and functional classes.

Main Methods:

• Development of two new artificial neural networks (ANNs) with multi-label classification capabilities.
• Training and testing the ANNs on protein residue sequences.
• Evaluation of classification performance using Area Under the Curve (AUC) metrics.

Main Results:

• The proposed ANNs achieved high accuracy in classifying protein sequences.
• Classification into 698 UniProt families yielded an AUC of 99.99%.
• Classification into 983 Gene Ontology classes resulted in an AUC of 99.45%.

Conclusions:

• The developed ANNs are highly effective for the multi-label classification of protein sequences.
• These models show significant potential for advancing biological sequence analysis and functional annotation.