Ab initio modeling led annotation suggests nucleic acid binding function for many DUFs.

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Rapid advancements in sequencing technology generate vast numbers of protein sequences, including many with unknown functions (Domains of Unknown Function, DUFs).
• Traditional homology-based annotation methods are insufficient for characterizing these novel protein domains.
• Structure-based function prediction offers a complementary approach to annotate protein functions.

Purpose of the Study:

• To computationally predict the function of short Domains of Unknown Function (DUFs).
• To identify DUFs likely involved in nucleic acid binding using ab initio modeling and structure-based approaches.
• To enhance the functional annotation of proteins within large sequence databases.

Main Methods:

• Ab initio modeling of short DUFs using the ROSETTA software suite.
• Screening of modeled DUFs for potential nucleic acid binding capabilities.
• Integration of supporting evidence from structure comparison, domain architecture, evolutionary relationships, genome context, and protein-protein interactions.

Main Results:

• Thirty-two DUFs were computationally predicted to possess nucleic acid binding functions.
• For most predicted DUFs, additional supporting evidence was identified through various biological data sources.
• These predictions facilitate the functional annotation of thousands of proteins containing these DUFs.

Conclusions:

• Structure-based ab initio modeling is effective for predicting the function of novel protein domains, particularly for nucleic acid binding.
• The integration of diverse biological data strengthens the confidence in predicted protein functions.
• This approach significantly contributes to the annotation of the proteome and understanding protein function in the era of big data.