PPI-PS

PPI-PS predicts protein-protein interactions by representing protein sequences as vectors of pairwise similarities computed via Smith-Waterman E-values and classifying protein pairs with a kernel support vector machine to distinguish interacted from non-interacted pairs, enabling investigation of interaction abnormalities such as those in neurological disorders.

Key Features:

• Shifting Window Technique: Applies a shifting window over concatenated training protein sequences to generate large subsequences of amino acids.
• Pairwise Similarity Calculation: Computes pairwise similarity scores between each protein sequence and the generated subsequences using the Smith-Waterman algorithm and represents them as E-values.
• Vector Representation: Encodes each protein sequence as a vector of pairwise similarity scores capturing interaction-relevant features.
• Kernel Matrix Construction: Uses the similarity-derived vectors to construct a kernel matrix for downstream classification.
• Support Vector Machine Classification: Employs support vector machines (SVMs) to classify pairs of proteins as interacted or non-interacted.

Scientific Applications:

• Protein–protein interaction prediction: Predicts binary interaction status between protein pairs from sequence-derived similarity vectors.
• Classification of interacting versus non-interacting pairs: Distinguishes interacted and non-interacted protein pairs using a kernel SVM applied to the kernel matrix.
• Study of neurological disorder-associated interactions: Supports analysis of interaction abnormalities implicated in neurological disorders by identifying altered interaction patterns.

Methodology:

A shifting window moves over concatenated training protein sequences to generate subsequences; each protein is compared to these subsequences with the Smith-Waterman algorithm to produce E-value pairwise similarities that form per-protein vectors, which are used to build a kernel matrix subsequently input to a support vector machine for classification of protein pairs.