Evolutionary profiles improve protein-protein interaction prediction from sequence.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Predicting protein-protein interactions (PPIs) from sequence is crucial for understanding cellular mechanisms.
• Existing sequence-based methods show limited performance for proteins dissimilar to training data.

Purpose of the Study:

• To develop a novel computational approach for predicting PPIs solely from protein sequences.
• To enhance prediction accuracy, particularly for evolutionarily distant or under-annotated proteins.

Main Methods:

• Utilized evolutionary profiles and profile-kernel support vector machines for PPI prediction.
• Incorporated gene expression data filtering to refine prediction accuracy for high-confidence interactions.

Main Results:

• The new method significantly outperformed existing state-of-the-art approaches.
• Performance gains were most pronounced for proteins with low sequence similarity to known interacting proteins.
• Gene expression filtering enhanced accuracy for a subset of highly reliable predictions.

Conclusions:

• The developed method offers a substantial improvement for predicting PPIs from sequence alone.
• This advancement is particularly impactful for annotating interactions of proteins lacking experimental data.
• The study provides a list of reliably predicted human PPIs, aiding future biological investigations.