Shared relationship analysis: ranking set cohesion and commonalities within a literature-derived relationship network

Area of Science:

• Bioinformatics
• Computational Biology
• Literature Mining

Background:

• Scientific research requires identifying commonalities among diverse objects like genes, phenotypes, chemicals, and diseases.
• The scientific literature is a rich source for uncovering relationships within heterogeneous object sets.

Purpose of the Study:

• To develop a method for identifying and evaluating commonalities and relationships among scientific objects.
• To quantitatively assess the cohesiveness of object sets and identify related items.

Main Methods:

• Constructing a network of related objects based on their co-occurrence in MEDLINE records.
• Querying object sets to find shared relationships and scoring their statistical relevance against a random network model.
• Utilizing Gene Ontology (GO) categories to demonstrate and validate the method's effectiveness.

Main Results:

• A network-based approach effectively identifies relationships between scientific objects.
• The method provides a quantitative measure of 'cohesiveness' for object sets.
• Demonstrated ability to identify and evaluate other objects for their 'cohesion' to a given set using GO categories.

Conclusions:

• This literature-mining approach offers a robust way to discover and quantify relationships among diverse scientific entities.
• The developed method aids in classifying, expanding, and understanding functional groupings of biological and chemical objects.
• Enables quantitative evaluation of object set cohesiveness and identification of novel related entities.