Species-specific model based on sequence and structural information for ubiquitination sites prediction

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Ubiquitination is a crucial post-translational modification regulating protein function in eukaryotes.
• Computational prediction of ubiquitination sites offers a cost-effective alternative to experimental methods.
• Existing models often struggle with limited structural data and species-specific variations.

Purpose of the Study:

• To develop a species-specific computational model for predicting ubiquitination sites.
• To address challenges posed by limited data and inter-species protein feature differences.
• To improve the accuracy and applicability of ubiquitination site prediction across diverse species.

Main Methods:

• Proposed a species-specific model (SSUbi) utilizing a capsule network architecture.
• Integrated multi-dimensional features from protein sequence and structural information.
• Employed convolution operations and channel attention mechanisms for feature extraction.

Main Results:

• SSUbi effectively integrates sequence and structural data for enhanced feature extraction.
• The model demonstrated improved prediction performance on species-specific datasets, especially those with small sample sizes.
• SSUbi outperformed existing computational models in predicting ubiquitination sites.

Conclusions:

• SSUbi provides an effective solution for species-specific ubiquitination site prediction.
• The model's ability to leverage limited data makes it valuable for understudied species.
• This approach advances computational methods in post-translational modification site prediction.