PPICT: an integrated deep neural network for predicting inter-protein PTM cross-talk

Area of Science:

• Biochemistry
• Computational Biology
• Systems Biology

Background:

• Post-translational modifications (PTMs) regulate protein function and signaling pathways.
• PTM cross-talk, involving modifications on residue pairs, is crucial but understudied.
• Existing methods lack integration of protein dynamics and protein-protein interaction (PPI) structural information for PTM cross-talk prediction.

Purpose of the Study:

• To develop a novel deep neural network model, PPICT, for predicting inter-protein PTM cross-talk.
• To integrate diverse biological data, including protein sequence, structure, dynamics, and PPI network topology.
• To enhance understanding of PTM functional mechanisms and signaling pathway regulation.

Main Methods:

• Developed PPICT, an integrated deep neural network combining protein sequence-structure-dynamics and PPI graph structural information.
• Utilized a heterogeneous feature combination network to leverage complex associations between protein evolutionary, biophysical, and interaction features.
• Trained and validated the model on comprehensive datasets to assess prediction performance.

Main Results:

• PPICT achieved a high prediction performance with an AUC value of 0.869, outperforming existing state-of-the-art methods.
• Identified that PTM cross-talk events are associated with residues exhibiting high co-evolution and allosteric regulation potential.
• Demonstrated PPICT's capability to identify potential PTM cross-talks involving modifying enzymes and their substrates.

Conclusions:

• PPICT is an effective computational tool for identifying PTM cross-talk at the proteome level.
• The findings provide insights into the structural and dynamic underpinnings of PTM cross-talk.
• This work facilitates a deeper understanding of PTMs in regulating signaling pathways and cellular functions.