3DCOOR-Kace: A 3-d spatial coordinates representation method for lysine acetylation site identification.

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Lysine acetylation (Kace) is a critical post-translational modification.
• Experimental determination of protein structures is time-consuming and resource-intensive, limiting large-scale analysis.
• Integrating structural information is crucial for enhancing computational models of Kace site prediction.

Purpose of the Study:

• To develop a novel spatial coordinate representation for amino acids to overcome experimental limitations.
• To construct complex networks using these coordinates for feature extraction.
• To propose an advanced computational model, 3DCOOR-Kace, for improved Kace site identification by fusing sequence and structure information.

Main Methods:

• A novel 3D spatial coordinate representation using a 2-order tensor based on defined property sequences.
• Construction of optimal complex networks from spatial coordinates to derive network-based features.
• Development of the 3DCOOR-Kace model, integrating sequence and structure data using DenseNet and Squeeze-and-Excitation layers.

Main Results:

• The proposed spatial coordinates effectively capture global biological information, outperforming traditional protein contact networks (PCN).
• The 3DCOOR-Kace model achieved a Matthews Correlation Coefficient (MCC) of 0.7358 on an independent testing set.
• 3DCOOR-Kace demonstrated superior performance compared to existing methods like MusiteDeep and TransPTM, with MCC improvements of 0.4261 and 0.1660, respectively.

Conclusions:

• The 3D spatial coordinate representation provides direct positional information, bypassing experimental limitations.
• The 3DCOOR-Kace model effectively integrates sequence and structural data for accurate Kace site prediction.
• This approach facilitates computational methods and biological function research by providing readily accessible positional data.