METSP: a maximum-entropy classifier based text mining tool for transporter-substrate identification with...

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Identifying transporter substrates is crucial for understanding transporter function, drug interactions, and metabolic pathways.
• Existing methods for substrate identification are limited, despite advancements in experimental techniques.
• Automated extraction of transporter-substrate pairs (TSPs) from biological literature remains a challenge.

Purpose of the Study:

• To develop and validate METSP, a novel maximum-entropy classifier for extracting TSPs from semistructured text.
• To apply METSP to a large corpus of human transporter annotations in UniProt to identify novel TSPs.
• To provide an efficient tool for researchers to discover transporter substrates and facilitate downstream applications.

Main Methods:

• Developed METSP, a maximum-entropy classification model.
• Utilized high-quality annotations from UniProt for training and validation.
• Applied METSP to 182,829 human transporter annotation sentences.
• Performed cross-validation experiments to assess precision and recall.

Main Results:

• METSP achieved high precision and recall in identifying TSPs.
• Identified 3942 sentences containing transporter and compound information from the UniProt corpus.
• Discovered 1547 high-confidence human TSPs, with 58.37% containing novel substrates not found in public databases.

Conclusions:

• METSP is the first efficient tool for extracting TSPs from semistructured UniProt annotation text.
• The identified novel TSPs can significantly aid in determining precise transporter substrates and drugs.
• This tool facilitates drug-target prediction, metabolic network reconstruction, and literature classification.