Functional annotation of creeping bentgrass protein sequences based on convolutional neural network

Area of Science:

• Plant Pathology
• Molecular Biology
• Bioinformatics

Background:

• Creeping bentgrass (Agrostis soionifera) exhibits poor disease resistance, a critical limitation for turfgrass applications.
• The induced systemic resistance (ISR) mechanism and its associated functional proteins in creeping bentgrass remain largely uncharacterized.
• Understanding ISR is crucial for developing disease-resistant turfgrass varieties.

Purpose of the Study:

• To identify functional proteins involved in the induced systemic resistance (ISR) mechanism in creeping bentgrass.
• To develop and apply a computational model for annotating uncharacterized protein sequences related to disease resistance.
• To enhance the understanding of molecular mechanisms underlying turfgrass defense responses.

Main Methods:

• Creeping bentgrass seedlings were treated and infected with Rhizoctonia solani to induce ISR.
• High-quality protein sequences were extracted from treated seedlings.
• A convolutional neural network (CNN) model was developed using Uniport database for annotating non-annotated protein sequences.

Main Results:

• A significant portion of protein sequences obtained from creeping bentgrass remained unannotated after standard database alignment.
• The developed CNN model successfully annotated non-annotated sequences, identifying proteins related to disease resistance and signal transduction.
• The model demonstrated good performance, particularly in controlling false positives.

Conclusions:

• A CNN-based prediction model effectively identified candidate proteins crucial for ISR in creeping bentgrass.
• This approach minimizes data loss and saves resources in future molecular biology research on turfgrass proteins.
• The study provides a valuable reference for sequence analysis in turfgrass disease resistance research.