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

## TL;DR

This paper identifies a recurring bias in network analysis used in ethnopharmacology and proposes a new AI-driven framework to improve its accuracy and relevance.

## Contribution

The paper introduces a bias-aware AI framework and roadmap to overcome homogeneity in network-based ethnopharmacology research.

## Key findings

- A narrow set of molecules like quercetin is repeatedly identified across studies, leading to homogeneity.
- Integrating multi-omics and experimental data reduces this homogeneity and improves biological relevance.
- A new framework is proposed to shift from database dependency to empirically driven, context-specific modeling.

## Abstract

Network analysis (NA) is a widely used computational tool for exploring the complex systems of interactions in ethnopharmacology, aiming to predict potential targets and generate mechanistic hypotheses. However, the predictive validity and biological relevance of its outputs are constrained by a pervasive methodological bottleneck: the recurrent identification of a narrow set of molecules—such as quercetin—across disparate natural products and diseases. Through a systematic analysis of 1,038 network-based studies, we establish “homogeneity” as a coherent, multi-level pattern, from “Flavonoid Centrality” to a “Hub-Target Core” and restricted “Canonical Pathways,” transcending specific remedies or diseases. We conceptualize this as a self-reinforcing “convergent discovery pipeline,” in which initial database biases are amplified by context-insensitive analytical approaches. Empirical evidence shows that integrating contextual experimental or multi-omics data mitigates homogeneity. To break this cycle and align network analysis more closely with pharmacological best practices, we propose an integrated framework that shifts from database dependency to empirically driven data acquisition, leverages bias-aware artificial intelligence for curation and prioritization, and advances dynamic, context-specific network modeling. This framework provides a clear roadmap to disrupt methodological inertia and steer network-based research in ethnopharmacology toward a more robust, diverse, and pharmacologically and clinically relevant future.