PMI-DB

PMI-DB provides a manually curated database of protein-metabolite interactions to support analysis and predictive modeling of protein–metabolite interactions.

Key Features:

• Size and composition: 49,785 manually curated entries comprising 23 small-molecule metabolites, 9,631 proteins, and four species.
• Interaction labels: Includes both interacting and non-interacting protein–metabolite pairs, providing explicit negative samples.
• Data source: Compiled in response to large-scale mass spectrometry analyses that uncovered protein–metabolite interaction data.
• Machine learning support: Balanced inclusion of positive and negative samples to support development and evaluation of predictive algorithms and models.
• Biological scope: Covers interactions relevant to metabolic enzymes, transcription factors, transporters, and membrane receptors.

Scientific Applications:

• Predictive model training and evaluation: Provides labeled positive and negative examples for training and validating machine learning classifiers of PMIs.
• Experimental planning: Enables reduction of unnecessary experiments by supplying known non-interacting pairs and interaction annotations.
• Functional regulation studies: Supports analysis of how metabolites regulate protein functions across metabolic enzymes, transcription factors, transporters, and membrane receptors.
• Cross-species analysis: Facilitates comparative studies using entries from the four represented species.

Methodology:

Entries were manually curated and compiled from data uncovered by large-scale mass spectrometry analyses and include both interacting and non-interacting protein–metabolite pairs.