Disease causality extraction based on lexical semantics and document-clause frequency from biomedical literature

Area of Science:

• Computational Biology
• Bioinformatics
• Medical Informatics

Background:

• Human disease network research is advancing, but current models often represent disease relationships as simple associations.
• This limitation hinders the identification of precursor diseases and their impact on subsequent conditions.
• A more sophisticated approach is needed to represent disease causality.

Purpose of the Study:

• To propose a novel causal disease network model.
• To implement disease causality inference through advanced text mining techniques.
• To overcome the limitations of existing association-based disease networks.

Main Methods:

• Developed a lexicon-based causality term strength scheme using lexicon analysis.
• Implemented a frequency-based causality strength scheme analyzing document and clause frequencies.
• Applied text mining to a large corpus of biomedical literature (6,617,833 PubMed articles).

Main Results:

• Constructed a causal disease network involving 195 diseases.
• Identified 1011 causal disease pairs among 149 diseases.
• The proposed method demonstrated superior performance, identifying 2.7 times more causalities and showing higher correlation compared to previous methods.

Conclusions:

• The novel method offers a cost- and time-efficient alternative to biological experiments for identifying disease causalities.
• This research advances text mining techniques by defining and utilizing "causality term strength."
• The causal disease network provides a more accurate and comprehensive understanding of disease interrelationships.