LSPpred

LSPpred predicts leaderless secretory proteins (LSPs) in plants, identifying proteins secreted without classical signal peptides to support secretome and proteomics analyses.

Key Features:

• Machine Learning-Based Prediction: Random forest classifiers trained on datasets derived from experimental observations are used to predict putative plant LSPs.
• Database Integration (LSPDB): Integration with the LSPDB plant protein database provides a repository of putative LSPs used in prediction and reference.
• Internal Validation and Accuracy Control: LSPpred and SPLpred modules are internally validated with false positive rates controlled at 5%, with SPLpred correctly identifying 3 out of 4 known examples and LSPpred correctly predicting all 4.

Scientific Applications:

• Facilitating Experimental Validation: Computational predictions of putative LSPs provide candidate lists to prioritize proteins for experimental validation.
• Refining Plant Proteomics and Secretome Analyses: Predicted LSPs aid in reducing contamination effects and improving interpretation of secretome datasets in plant proteomics workflows.

Methodology:

Criteria for identifying candidate LSPs were established from experimental observations and used to train random forest classifiers on datasets linked to the LSPDB, and the resulting modules were internally validated to control false positives at 5%.