SimText

SimText performs text mining and similarity analysis to quantify and visualize relationships among biomedical entities, including genes, diseases, and experiments, using PubMed-derived text.

Key Features:

• Data collection from PubMed: Retrieves textual data from PubMed including abstracts and related textual content for specified biomedical entities.
• Word extraction / text mining: Extracts words and terms from retrieved texts using diverse text mining approaches to represent entity-associated vocabulary.
• Similarity matrix construction: Constructs similarity matrices that capture relationships among entities based on shared or related textual features.
• Unsupervised learning analysis: Applies unsupervised learning techniques to explore patterns and relationships without predefined labels.
• Clustering and dimensionality reduction: Uses clustering and dimensionality reduction methods to identify groupings and low-dimensional representations of entity similarities.
• Visualization outputs: Produces visual representations of similarity relationships and analysis results to facilitate interpretation.

Scientific Applications:

• Comparative analysis of biomedical entities: Enables comparison of genes, diseases, and experiments based on literature-derived textual similarity.
• Literature-driven similarity mapping: Supports mapping of related entities and hypothesis generation from PubMed abstracts.
• Large-scale studies and profiling: Applicable to large datasets such as genome-wide association studies and comprehensive disease profiling.

Methodology:

Collect text from PubMed, extract words from abstracts, construct similarity matrices from extracted terms, and analyze matrices using unsupervised learning including clustering and dimensionality reduction methods.