Neurodegenerative diseases such as Alzheimer’s disease (AD), Parkinson’s disease (PD), and Huntington’s disease (HD) are progressive disorders with limited disease‑modifying therapies. A shared pathological feature across these conditions is accumulation of misfolded proteins (amyloid‑β and tau in AD, α‑synuclein in PD, mutant huntingtin in HD) that contribute to synaptic dysfunction, neuroinflammation, mitochondrial impairment, oxidative stress and neuronal loss. Given limitations of current symptomatic treatments and the multi‑factorial nature of neurodegeneration, plant‑derived multi‑target compounds are of interest. Centella asiatica, an edible and medicinal herb, has reported antioxidant, anti‑inflammatory, mitochondrial‑protective and neuroprotective properties; several of its compounds reportedly cross the blood–brain barrier. However, the molecular targets and pathways mediating potential benefits in AD, PD and HD remain incompletely defined.
The authors implemented an integrative in silico workflow combining network pharmacology, transcriptomic analyses, machine learning and molecular docking to prioritize candidate C. asiatica constituents and disease‑relevant targets. Candidate phytochemicals from C. asiatica were filtered using predefined criteria for drug‑likeness, gastrointestinal absorption and blood–brain barrier permeability. Predicted compound targets were compiled and merged with disease‑associated gene sets obtained by disease‑gene mining for AD, PD and HD. Public transcriptomic datasets from the Gene Expression Omnibus were used to incorporate differential expression and to refine target prioritization; specific GEO accession numbers analyzed were reported. Machine learning and network centrality metrics were applied to identify hub genes. Receiver operating characteristic (ROC) analyses in independent datasets assessed discriminatory performance of prioritized genes. Finally, molecular docking (reported as Vina scores) evaluated putative physical interactions between selected compounds and hub targets.
Sixteen C. asiatica constituents passed the drug‑likeness, gastrointestinal absorption and blood–brain barrier filters and were associated with 370 unique predicted protein targets. Disease‑gene mining returned 983 AD‑associated genes, 1,103 PD‑associated genes and 3,316 HD‑associated genes. By integrating compound targets with disease gene sets and transcriptomic evidence, the analysis prioritized multiple hub genes for each disorder.
For PD, five hub genes were highlighted: CCKAR, MAPK8, PSEN2, SLC6A3, and TH. For AD, four hub genes were prioritized: APP, PGK1, PIK3CA, and TTR. For HD, four hub genes were identified: CHRND, HSP90AA1, PRKCQ, and TH. These genes emerged from combined analyses of predicted compound interactions, disease association and transcriptomic differential expression.
ROC analyses performed on independent datasets were reported to provide additional support for the discriminatory performance of the prioritized genes, suggesting potential relevance for disease classification or biomarker candidacy. Molecular docking produced favourable predicted Vina docking scores and structurally plausible interactions between selected C. asiatica compounds and some hub targets, supporting the feasibility of compound–target binding inferred by the network analyses.
Gene‑set enrichment analyses contextualized the prioritized targets within biological processes and signalling cascades relevant to neurodegeneration. Enriched themes included neurotransmitter signalling, cyclic AMP (cAMP) and calcium signalling pathways, MAPK‑related responses, and regulation of inflammatory processes. These pathway associations align with known mechanisms implicated in AD, PD and HD pathophysiology and with previously reported neuroprotective actions of some C. asiatica constituents.
The authors report that ROC curve analyses in independent transcriptomic datasets supported the discriminatory capacity of the prioritized genes, although specific performance metrics are presented in the source article. Molecular docking results identified predicted binding poses and Vina scores consistent with structurally plausible interactions between selected phytochemicals and hub proteins. Detailed docking scores, interaction residues and compound‑target pairings were provided in the original figures and supplementary files.
This study demonstrates a multi‑layered computational approach to prioritize candidate bioactive constituents from a medicinal food‑herb and to nominate putative molecular targets and pathways in three major neurodegenerative disorders. By filtering for blood–brain barrier permeability and integrating public transcriptomic evidence, the analysis aimed to increase biological plausibility for CNS activity. The prioritized hub genes and enriched pathways converge on neurotransmission, signal transduction and inflammatory modulation—processes previously implicated in neurodegeneration and in reported pharmacology of C. asiatica extracts.
The authors note that these results are hypothesis‑generating and emphasize the need for biochemical, cellular and in vivo validation to confirm compound bioactivity, target engagement, dose‑response relationships and therapeutic effects. They also make available processed gene‑expression matrices, differential‑expression results and anonymized analysis code in Zenodo to promote transparency and reproducibility.
An integrative computational pipeline combining network pharmacology, transcriptomics, machine learning and molecular docking prioritized 16 Centella asiatica constituents, 370 predicted targets, and disorder‑specific hub genes and pathways in AD, PD and HD. The study provides a prioritized list of candidate compound–target relationships and mechanistic hypotheses to guide subsequent experimental validation.
No new experimental raw data were generated in the study. The transcriptomic datasets analyzed were publicly available from GEO under accession numbers reported in the source. Processed matrices, differential‑expression results and analysis code were deposited in Zenodo at the DOI cited in the original article. Funding was provided by the Youth Guidance Project of the Guizhou Provincial Basic Research Program (Natural Science). The authors declared no competing interests.