Advancing biological taxonomy in the AI era: deep learning applications, challenges, and future directions

Area of Science:

• Biological taxonomy
• Bioinformatics
• Artificial Intelligence

Background:

• Biological taxonomy has evolved through distinct technology-driven eras: morphology, molecular analysis, and now artificial intelligence (AI).
• Each new technology has augmented, not replaced, previous methods, expanding the scope of taxonomic research.
• The current AI-driven stage, particularly deep learning, shows transformative potential.

Purpose of the Study:

• To review the progress of biological taxonomy through its technology-driven eras.
• To elucidate the impact of deep learning and foundation models on taxonomic classification and trait elucidation.
• To discuss the challenges and opportunities in integrating AI into biological taxonomy.

Main Methods:

• Review of technological advancements in biological taxonomy.
• Analysis of deep learning applications in image-based, bioacoustics-based, and genetic sequence-based classification.
• Exploration of foundation models in linking genomic data to phenotype and ecology.

Main Results:

• Deep learning has significantly impacted biological classification across various data types (images, sounds, sequences).
• Foundation models are beginning to connect genomic variations with protein structure, phenotype, and ecological niche.
• AI offers a potential step-change in taxonomy through integrated, causality-aware models.

Conclusions:

• AI, especially deep learning and foundation models, presents a paradigm shift in biological taxonomy.
• Key challenges include data quality, algorithmic robustness, model transparency, and standardization.
• Taxonomists' expertise is crucial for guiding AI development and navigating the evolution of taxonomic concepts.