DeepCSO

DeepCSO predicts cysteine S-sulphenylation (CSO) sites in proteins to identify post-translational modification loci that influence protein function and signaling.

Key Features:

• Three LSTM models: Implements three long short-term memory (LSTM) models comprising two species-specific models and one general model that integrates multi-species data.
• LSTM with word-embedding encoding (LSTMWE): Employs LSTMWE encoding and modelling, reported to outperform traditional machine-learning and other deep-learning approaches with area under the ROC curve values of 0.82–0.85.
• Expanded multi-species dataset: Trains models on an expanded dataset of identified CSO sites across several species, enabling broader generalization beyond tools limited to Homo sapiens data.

Scientific Applications:

• Cross-species comparative analysis: Enables analysis of CSO characteristics across multiple species for comparative proteomics studies.
• Protein function regulation research: Supports investigation of how cysteine S-sulphenylation modulates protein function and cellular signaling pathways.
• Proteome-wide CSO prediction: Facilitates prediction of potential CSO sites within proteins to aid studies of post-translational modification patterns at proteomic scale.

Methodology:

Uses long short-term memory (LSTM) networks with word-embedding encoding (LSTMWE), implementing three models (two species-specific, one general) trained on an expanded multi-species CSO dataset, with reported AUC 0.82–0.85.