Rethinking network analysis in ethnopharmacology: a multi-omics and AI roadmap to overcome conceptual and...

Area of Science:

• Computational biology
• Pharmacology
• Ethnobotany

Background:

• Network analysis (NA) is crucial for ethnopharmacology, predicting targets and hypotheses.
• A key limitation is the recurrent identification of a narrow set of molecules (e.g., quercetin) across diverse studies.
• This homogeneity stems from database biases amplified by analytical methods, creating a 'convergent discovery pipeline'.

Purpose of the Study:

• To systematically analyze the 'homogeneity' pattern in network-based ethnopharmacology studies.
• To conceptualize the mechanisms driving this homogeneity.
• To propose a novel framework to enhance the diversity and relevance of NA in ethnopharmacology.

Main Methods:

• Systematic analysis of 1,038 network-based ethnopharmacology studies.
• Identification and characterization of the 'homogeneity' pattern across multiple levels (Flavonoid Centrality, Hub-Target Core, Canonical Pathways).
• Evaluation of methods to mitigate homogeneity, including integration of contextual experimental and multi-omics data.

Main Results:

• A multi-level 'homogeneity' pattern was established, characterized by 'Flavonoid Centrality,' a 'Hub-Target Core,' and restricted 'Canonical Pathways.'
• This pattern represents a self-reinforcing 'convergent discovery pipeline' driven by database biases and insensitive analytical approaches.
• Integrating contextual experimental or multi-omics data was shown to effectively mitigate homogeneity.

Conclusions:

• The pervasive homogeneity in NA hinders the predictive validity and biological relevance of ethnopharmacological research.
• A shift from database dependency to empirically driven data acquisition is necessary.
• An integrated framework utilizing bias-aware AI and dynamic network modeling can foster more robust, diverse, and clinically relevant ethnopharmacological discoveries.