Biomaterials text mining: A hands-on comparative study of methods on polydioxanone biocompatibility

Area of Science:

• Biomaterials Science
• Computational Biology
• Information Science

Background:

• Scientific information extraction is crucial but largely manual and time-consuming.
• Text Mining tools (TMTs) offer automated extraction but are underutilized in biomaterials.
• Polydioxanone biocompatibility research, a niche area, serves as a case study.

Purpose of the Study:

• To compare the efficacy of various TMTs for extracting information from biomaterials abstracts.
• To evaluate TMTs' ability to identify themes, track topic evolution, and discover applications in biomaterials literature.
• To assess the potential of Natural Language Processing (NLP) and domain-specific tools in biomaterials data organization.

Main Methods:

• Tested multiple TMTs, including machine learning, statistical analysis, MeSH indexing, and Named Entity Recognition (NER) tools.
• Focused on abstracts related to the biocompatibility of polydioxanone.
• Compared TMT output with manual review of systematic reviews and meta-analyses.

Main Results:

• TMTs demonstrated high efficiency in mapping biomaterials texts and providing current information.
• Tools successfully identified dominant themes, term evolution, and key medical applications.
• Ambiguity in biomaterials nomenclature presents a significant challenge for automated extraction.

Conclusions:

• TMTs are powerful tools for rapidly extracting and organizing valuable information from biomaterials literature.
• NLP and domain-specific tools hold significant potential for advancing biomaterials data management.
• Addressing nomenclature ambiguity is essential for improving future text mining efforts in this field.