POOE: predicting oomycete effectors based on a pre-trained large protein language model.

Area of Science:

• Plant Pathology
• Bioinformatics
• Molecular Biology

Background:

• Oomycetes cause devastating plant diseases by secreting effector proteins.
• Accurate identification of oomycete effectors is vital for understanding plant immunity.
• Existing prediction tools for oomycete effectors have limitations in performance.

Purpose of the Study:

• To develop a highly accurate bioinformatics tool for predicting oomycete effectors.
• To leverage large protein language models for improved effector prediction.
• To accelerate the discovery and functional analysis of oomycete effectors.

Main Methods:

• Utilized sequence embeddings from a pre-trained protein language model (ProtTrans).
• Developed a Support Vector Machine (SVM)-based prediction model named POOE.
• Validated performance using fivefold cross-validation and an independent test set.

Main Results:

• POOE achieved high accuracy (0.874) and AUC (0.893) in cross-validation.
• The model significantly outperformed existing prediction methods and encoding schemes.
• Consistent high performance was observed on the independent test set.

Conclusions:

• POOE demonstrates a powerful new approach for oomycete effector prediction.
• The use of ProtTrans effectively captures protein semantic information for improved prediction.
• POOE is expected to accelerate research into plant-pathogen interactions and effector functions.