JMIR Publications

ABSTRACT

Background:

Natural language processing has long been applied in various application on bio-medical knowledge inference and discovery. A routine bio-medical NLP method performed named entity recognition to extract bio-terms, e.g., gene, chemical, mutation, in abundant text resources and investigated the associations between the entities. Among the analysis, terms enrichment was a classic one.

Objective:

To examine whether Inverse Pathway Frequency (IPF), a novel metric, is effective in the text-mined terms enrichment and test robustness of the newly proposed metric through comparison between IPF and traditional evaluation metrics.

Methods:

First, use various bio-medical text mining strategy to investigate drug-related genes set. Second, design novel metric for pathway enrichment of text mined genes in the support of drug repurposing. In this case, seven novel Inverse Pathway Frequency (IPF) metrics are proposed and under comparison with traditional p-value. Finally, carry on a case study to show the effectiveness of the metrics and as well the promising application of text mining pipeline for drug repurposing.

Results:

A known gene-pathway association was re-discovered with the implementation of the IPF metric, and it was visualized by Cytoscape.

Conclusions:

The results showed the effectiveness of IPF metric, as well as the promising application of NLP on bio-medical knowledge discovery. Clinical Trial: None