Abstract
Network pharmacology nominates targets but cannot show that they are altered in the relevant cell, distinguishable from a hypothesis-free analysis of the same data, or linked to disease. Taking the candidates nominated for Yangxin Decoction in dilated cardiomyopathy (DCM) and viral myocarditis, we asked how many survive independent functional-genomic testing. Few did. A direct-evidence intersection returned 61 genes, but across four bulk cardiac datasets the set was differentially expressed no more often than the transcriptome-wide background in any human cohort, and AKT1 was significant in none. Donor-pseudobulk single-nucleus RNA sequencing of 52 DCM and 18 control donors showed higher AKT1 (log2 fold change 0.351, adjusted P = 3.32 × 10- 4), PIK3CA and GSK3B and lower STAT3 in cardiomyocytes; no fold-change threshold was applied, and the AKT1 difference is small. These differences were robust to nucleus depth, sex, age, ventricle and DCM genotype; in an independent single-nucleus DCM cohort AKT1 (0.530) and STAT3 (- 0.771) replicated at FDR - 3), so cellular dilution does not explain the negative bulk-microarray result. Benchmarked against an unbiased genome-wide analysis of the same cardiomyocyte comparison, the nominated 61-gene set was not enriched (46.7% versus a 41.3% background, P = 0.281) and AKT1 ranked 2,182 of 14,586. A prespecified screen found no cis-eQTL instrument for AKT1; a single whole-blood MTOR Wald ratio was not supported by colocalization. These are cell-type-resolved transcriptional differences in dilated cardiomyopathy, not evidence that Yangxin Decoction engages PI3K-AKT signalling and not a validated therapeutic mechanism.