Challenges and limitations

Despite the promising mechanistic insights offered by these networkpharmacology-based studies on herbal extracts in asthma, several conceptual and practical challenges remain to be addressed, both in methodology and in clinical translation (Figure 3).

Figure 3.: Challenges in network pharmacology

Network pharmacology, despite its conceptual strengths, faces several practical and methodological challenges that limit the robustness and clinical translatability of its predictions. One major issue is the inconsistency and incompleteness of underlying databases. Because the discipline depends heavily on network and target databases for primary data, outcomes can differ substantially between studies or fail to align with experimental observations, especially when poorly curated or outdated entries are used [42].

Another critical concern is the quality and safety assessment of herbal medicines. The concentrations of active and potentially toxic constituents can vary widely across batches, regions and preparation methods, and many studies on traditional herbal formulations do not follow standardized quality-control or safety protocols. This variability complicates clinical evaluation, hinders reproducibility and weakens the reliability of networkbased predictions. Furthermore, network models often overlook systemic physiological integration, focusing on intracellular molecular networks while neglecting interorgan communication and broader regulatory systems such as the neural, immune and endocrine axes, which play central roles in diseases like asthma. The compatibility and standardization of herbal extracts pose additional hurdles. Achieving pharmaceutical-grade quality and batch-to-batch reproducibility with complex mixtures is difficult, undermining both experimental consistency and clinical relevance. At the mechanistic level, network models may also fail to account for resistance mechanisms, such as efflux transporters like Pglycoprotein, that can significantly modulate drug exposure and override predicted pharmacological effects. Moreover, many networkpharmacology studies remain heavily reliant on in silico outputs without adequate experimental validation; well-designed preclinical and clinical investigations are therefore essential to confirm predicted relationships between gene expression changes and actual physiological or therapeutic outcomes [56].

Methodological limitations also arise from the screening tools and criteria commonly applied. Traditional rules such as Lipinski’s Rule of Five are tailored to smallmolecule synthetic drugs and may inappropriately exclude promising herbal compounds that fall outside these narrow physicochemical ranges [56]. At the same time, current databases often ignore the actual content and dose of individual components in herbal preparations, even though efficacy depends on whether bioactive constituents reach therapeutic concentrations in target tissues. The chemical composition of herbal medicines can change during preparation (e.g. during decoction, influenced by time, temperature and solvent), but such variability is not reflected in existing network databases or analyses. Finally, an overreliance on target number can distort the identification of key active components, as methods that prioritize compounds with many predicted targets may overlook agents with fewer but more potent and functionally relevant interactions [50].