Spotlite

Spotlite integrates APMS scoring algorithms and indirect evidence to improve prediction of protein-protein interactions from affinity purification–mass spectrometry (APMS) data.

Key Features:

• Integration of Proven Scoring Approaches: Incorporates SAINT, CompPASS, and HGSCore and augments their outputs with mRNA coexpression, gene ontology annotations, domain-domain binding affinities, and homologous protein interactions.
• Logistic Regression Classifier: Combines scores from multiple APMS scoring algorithms with indirect features using a logistic regression model to improve interaction classification, producing an average 16% increase in area under the ROC curve across five APMS datasets.
• Complementary Scoring Aggregation: Leverages complementary classification accuracies from different scoring strategies to prioritize genuine protein-protein interactions and refine candidate interaction lists.
• Contaminant Pruning and Annotation: Uses integrated indirect evidence to prune contaminants from APMS datasets and annotate candidate interactions for downstream analysis.

Scientific Applications:

• Protein-Protein Interaction Discovery: Improves identification of bona fide interactions within APMS datasets.
• Interaction Network Construction and Functional Interpretation: Supports building interaction networks and interpreting functional relationships using integrated scoring and annotation.
• Disease-Relevant Interaction Analysis: Enables focused analyses that can reveal interaction details relevant to disease biology, as exemplified by studies of the KEAP1 E3 ubiquitin ligase.

Methodology:

Performs comparative analysis of SAINT, CompPASS, and HGSCore, integrates their outputs with indirect data sources (mRNA coexpression, gene ontology, domain-domain affinities, homologous interactions), and applies a logistic regression classifier to improve APMS interaction prediction, validated by AUC increases across five datasets.