Sitetack: A Deep Learning Model that Improves PTM Prediction by Using Known PTMs.

Area of Science:

• Proteomics and Bioinformatics
• Computational Biology
• Machine Learning in Biology

Background:

• Post-translational modifications (PTMs) are crucial for proteome diversity, organismal life, and therapeutic development.
• Predicting PTM locations using deep learning shows promise but is hindered by dataset limitations and analysis challenges.
• Existing sequence-based deep learning models for PTM prediction have limitations in accuracy and scope.

Purpose of the Study:

• To evaluate the impact of incorporating known PTM sites into sequence-based deep learning algorithms for improved PTM location prediction.
• To investigate whether knowledge of one PTM's location can enhance the prediction of other PTMs.

Main Methods:

• PTM locations were encoded as distinct amino acid residues within protein sequences.
• Protein sequences were then encoded using word embedding techniques.
• A convolutional neural network (CNN) was employed to predict the probability of a modification at each site.

Main Results:

• The model's performance was comparable to existing methods when known PTMs were not explicitly labeled.
• Labeling known PTM sites led to a significant improvement in prediction accuracy compared to current models.
• The model demonstrated that knowledge of existing PTM locations can enhance the predictability of other PTMs.

Conclusions:

• The findings underscore the critical role of PTMs in the sequential installation of additional PTMs on proteins.
• Including information about known PTM locations is a valuable strategy for enhancing the performance of proteomic machine learning algorithms.
• This approach offers a pathway to more accurate PTM prediction, advancing proteomic research and therapeutic applications.