Alkemio

Alkemio identifies and ranks chemicals associated with user-defined biomedical topics by text-mining gene-related PubMed® abstracts and statistically prioritizing candidate chemicals.

Key Features:

• Automated Text Mining: Processes millions of PubMed® articles and gene-related abstracts to detect mentions of chemicals, including drugs.
• Naïve Bayesian Classifier: Uses a naïve Bayesian classifier to predict the relatedness of chemicals to user-defined query topics.
• Ranking by P-values: Ranks chemicals using P-values derived from random simulations.
• High Performance Metrics: Benchmarking on seven human pathways yielded areas under the receiver operating characteristic curves ranging from 73.6% to 94.5%.
• Comparative Advantage: Exhibits superior precision and recall versus existing tools for identifying chemicals associated with eight diseases, particularly among the top 10 candidate chemicals.

Scientific Applications:

• Drug Discovery: Identifies potential drug candidates related to specific diseases or pathways.
• Pathway Analysis: Enables exploration of chemical involvement in biological pathways.
• Disease Mechanism Studies: Supports identification of chemical–disease associations relevant to disease mechanisms and therapeutic targets.

Methodology:

Retrieval of gene-related PubMed® articles, application of text-mining to article content followed by classification with a naïve Bayesian model, and ranking of chemicals using P-values from random simulations.