A Neuro-Symbolic Bioinformatics Framework for Unlocking Chordate Physiological Dark Data and Validating Allometric...

Area of Science:

• Bioinformatics
• Macroecology
• Computational Biology

Background:

• Animal functional trait data is crucial for macroecology but is often trapped in unstructured literature.
• Manual data extraction is slow, and existing AI struggles with biological tables and numerical precision.

Purpose of the Study:

• To develop an efficient and accurate method for extracting biophysical trait data from scientific literature.
• To construct a high-fidelity physiological database for chordate species.

Main Methods:

• A multimodal neuro-symbolic AI framework combining visual-language perception and code-based reasoning was employed.
• The framework reconstructs document layouts and uses isolated programming for accurate biostatistical calculations.
• Literature spanning 117 years was mined to build the database.

Main Results:

• A high-fidelity physiological database for 1632 chordate species was constructed.
• The method achieved a macro-averaged F1 score of 0.935 for extracting biophysical fields.
• Extracted data reproduced known allometric scaling relationships and showed strong concordance with existing databases.

Conclusions:

• The study presents a reproducible bioinformatics pipeline for extracting and standardizing physiological data from literature.
• This approach minimizes extraction errors and provides a valuable resource for building physiology-oriented trait databases.
• The validated pipeline supports scalable data construction for macroecological studies.