Structure-aware annotation of leucine-rich repeat domains.

Area of Science:

• Structural bioinformatics
• Computational biology
• Plant innate immunity

Background:

• Protein domain annotation traditionally relies on sequence motifs, which often leads to errors in domain boundary and motif identification.
• Existing sequence-based methods lack crucial structural information, limiting their accuracy.
• Deep learning-based protein structure prediction offers new opportunities to enhance domain annotation by incorporating structural geometry.

Purpose of the Study:

• To develop novel dimensionality reduction methods for accurate annotation of repeat units within Leucine Rich Repeat (LRR) solenoid domains.
• To improve upon existing machine learning-based domain annotation tools by integrating structural information.
• To enable automated detection of hairpin loops and structural anomalies in LRR solenoids.

Main Methods:

• Application of dimensionality reduction techniques to analyze protein structures.
• Development of methods to annotate repeat units in Leucine Rich Repeat (LRR) solenoid domains.
• Integration of predicted protein structure geometry into annotation pipelines.

Main Results:

• The developed methods successfully corrected errors made by existing machine learning-based annotation tools.
• Automated detection of hairpin loops and structural anomalies within LRR solenoids was achieved.
• The approach was applied to 117 predicted structures of LRR-containing proteins in Arabidopsis thaliana, with validation against 172 manually annotated LRR domains.

Conclusions:

• Incorporating protein structure geometry significantly improves the accuracy of domain annotation compared to sequence-only methods.
• The new methods provide a robust framework for analyzing LRR domains and identifying structural variations.
• This work advances the understanding of LRR-containing proteins involved in plant innate immunity.