POOE

POOE predicts oomycete effector proteins from amino acid sequences using ProtTrans-derived sequence embeddings and a support vector machine to support studies of plant–pathogen interactions.

Key Features:

• Advanced Methodology: Generates sequence embeddings from the ProtTrans pre-trained protein language model to capture semantic information in protein sequences.
• Machine Learning Approach: Implements a Support Vector Machine (SVM)-based classifier optimized for effector prediction.
• Performance Metrics: Achieved an AUPRC of 0.804, an AUROC of 0.893, accuracy 0.874, precision 0.777, recall 0.684, and specificity 0.936 in fivefold cross-validation, indicating superior performance relative to existing oomycete effector predictors.
• Independent Validation: Demonstrated consistent performance on independent test sets, outperforming other commonly used sequence encoding schemes and machine learning algorithms.

Scientific Applications:

• Plant–pathogen interaction research: Facilitates identification of potential oomycete effector proteins to study mechanisms of host manipulation.
• Functional characterization: Guides selection of candidate effectors for experimental validation and functional studies.
• Disease management research: Informs development of strategies for managing oomycete-related plant diseases in agricultural systems.

Methodology:

Generates ProtTrans-derived sequence embeddings and trains an SVM classifier evaluated by fivefold cross-validation and independent test sets.