Prediction of Protein-Protein Interaction via co-occurring Aligned Pattern Clusters

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Protein-protein interactions (PPIs) are crucial for cellular mechanisms.
• Experimental PPI identification is costly and inefficient.
• Sequence-based prediction offers a cost-effective alternative.

Purpose of the Study:

• To develop a novel, interpretable, and effective sequence-based method for PPI prediction.
• To overcome limitations of existing feature-based and kernel-based methods.
• To identify biologically meaningful features driving PPIs.

Main Methods:

• Developed WeMine-P2P to discover conserved, variable-length sequence patterns (APCs).
• Paired APCs into co-occurring APC (cAPC) pairs to compute cAPC-PPI scores.
• Constructed feature vectors from cAPC scores for PPI prediction.

Main Results:

• WeMine-P2P outperformed PIPE2 in PPI prediction.
• Achieved performance comparable to SVM-based methods, especially for novel sequences.
• Identified biologically interpretable features, highlighting compositional bias regions.

Conclusions:

• WeMine-P2P provides an interpretable and effective approach to PPI prediction.
• The method demonstrates potential for predicting other biosequence interactions, like Protein-DNA interactions.