LAITOR4HPC: A text mining pipeline based on HPC for building interaction networks

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• The exponential growth of scientific literature necessitates advanced text mining tools for biological network construction.
• Existing tools like PESCADOR face limitations in processing large-scale data efficiently.
• LAITOR4HPC is developed to overcome these limitations by leveraging high-performance computing (HPC).

Purpose of the Study:

• To present LAITOR4HPC, an HPC-oriented version of the LAITOR text mining tool.
• To enable rapid analysis of unlimited biological abstracts for network enrichment and construction.
• To highlight the tool's capability in identifying research gaps and enhancing biological pathways.

Main Methods:

• Utilizing parallel computing on HPC infrastructure for accelerated abstract analysis.
• Processing the entire MEDLINE abstract collection up to June 2017.
• Demonstrating utility through three distinct case studies: single-organism co-occurrence, multi-organism gene family analysis, and pathway enrichment.

Main Results:

• Complete MEDLINE abstract analysis (until June 2017) completed in 6 days, a significant improvement over the original implementation.
• Case studies successfully retrieved extensive protein-protein co-occurrences, analyzed gene families across species, and enriched the defensin pathway in Arabidopsis thaliana.
• Identified 15,788 proteins and 7894 co-occurrences for soybean, and highlighted key species like Arabidopsis thaliana and Zea mays in biotic stress analysis.

Conclusions:

• LAITOR4HPC facilitates efficient text mining for constructing biological networks from large datasets like MEDLINE abstracts.
• The tool's performance and data capacity are scalable based on available HPC resources.
• LAITOR4HPC offers flexibility for diverse research needs, enabling organism-specific, subject-specific, or pathway-specific analyses with reliable interaction classification.