Integrated multi-omics and deep learning analysis reveals neurotransmitter metabolism regulatory mechanisms of
Macao Wan1, Rangyuzhen Cai2, Xusheng Zhang3
1Special Operations Department, The 940 Hospital of the Joint Logistics Support Force of Chinese PLA, Lanzhou, Gansu, China.
Background:
Tianwang Buxin Dan is a classical Traditional Chinese Medicine formula with documented clinical use in treating neuropsychiatric disorders, yet its molecular mechanisms remain incompletely understood.
Methods:
We developed an integrated analytical framework combining transcriptomic and metabolomic profiling with deep learning to investigate the neurotransmitter metabolism regulatory mechanisms of Tianwang Buxin Dan. Data were collected from a para-chlorophenylalanine (PCPA)- induced insomnia rat model following formula intervention. A multi-omics feature fusion strategy incorporating autoencoder-based dimensionality reduction and cross-modal attention mechanisms was implemented to address data heterogeneity.
Results:
A total of 1,847 differentially expressed genes and 286 differential metabolites were identified. The constructed deep neural network achieved 91.2% classification accuracy with an AUC of 0.956 in five-fold cross-validation, and permutation testing confirmed that performance was significantly above chance (p < 0.001). Ablation experiments demonstrated that integrated multiomics outperformed single-omics models. Tryptophan hydroxylase 2 (TPH2) upregulation and monoamine oxidase A (MAO-A) suppression were identified as key features and partially validated by qPCR and Western blot.
Discussion:
Tianwang Buxin Dan may modulate neurotransmitter metabolism through coordinated regulation of biosynthetic and catabolic pathways. A component-target-pathway regulatory network identified 47 key molecular targets interconnected through 156 functional associations. This work provides a computational framework applicable to mechanism studies of other compound TCM formulations.
