On the importance of comprehensible classification models for protein function prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Current protein function prediction research prioritizes predictive accuracy.
• This focus often overlooks the critical aspects of knowledge validation and interpretation.
• This can hinder the generation of biologically meaningful insights and hypotheses.

Purpose of the Study:

• To critically evaluate the prevailing approach in protein function prediction.
• To introduce a new perspective emphasizing model comprehensibility alongside predictive accuracy.
• To highlight the importance of interpretable models for biological discovery.

Main Methods:

• The paper argues for the development of comprehensible protein function prediction models.
• It discusses the advantages of such models, including increased biologist confidence and error detection.
• A critical review of various knowledge representations supporting comprehensible models is presented.

Main Results:

• Comprehensible models enhance biologist confidence in predictions.
• Interpretable models facilitate the discovery of new biological insights and hypotheses.
• Comprehensible models aid in identifying potential errors within the data.

Conclusions:

• Protein function prediction should prioritize both accuracy and comprehensibility.
• Interpretable models are crucial for advancing biological understanding and hypothesis generation.
• The choice of knowledge representation significantly impacts model comprehensibility.