Journal: bioRxiv

Article Title: Temporal AI model predicts drivers of cell state trajectories across human aging

doi: 10.64898/2026.03.30.715396

Figure Lengend Snippet: ( A ) Predicted impact of in silico inhibition vs. slope of expression change across human aging for genes in cardiac fibroblasts. ( B ) Concordant genes whose inhibition was predicted to be rejuvenating were upregulated across aging, and vice versa. ( C ) Predicted impact of in silico inhibition of genes across the indicated cardiac cell types. ( D ) Beta-gal staining and cell counts quantifying senescence of primary human cardiac capillary endothelial cells in response to predicted pro-aging perturbation of ZBTB16 inhibition. *p<0.05, Wilcoxon rank sums, n=4. ( E ) Predicted impact of in silico perturbation and expression change across human aging and in telomere-shortened mice in each cardiac cell type for top predicted cardiomyocyte age-modulating targets. ( F ) PCA plot and ( G ) differential gene expression heatmap of transcriptional response (measured by bulk RNA sequencing) to predicted pro-aging perturbations (AAV overexpression of indicated genes vs. GFP) in human iPSC-derived cardiomyocytes. p<0.05, Wald test with BH correction, n=4. ( H ) Gene set enrichment of genes differentially expressed in response to all predicted pro-aging perturbations in human iPSC-derived cardiomyocytes (hypergeometric test with g:Set Counts and Sizes (g:SCS) correction). ( I ) Slowed calcium cycle kinetics and ( J ) rhythm irregularities in response to predicted pro-aging perturbations (AAV overexpression of indicated genes vs. GFP) in human iPSC-derived cardiomyocytes. *p<0.05, Wilcoxon rank sums with BH correction. Rhythm n=10, time to peak n=180, 111, 139, 37, 237, 150, 191 (bars left to right), decay n=169, 30, 86, 11, 213, 105, 172 (bars left to right). ( K ) Differential gene expression heatmap of transcriptional response (measured by bulk RNA sequencing) to predicted pro-aging perturbations (AAV overexpression of indicated genes vs. GFP) in human iPSC-derived cardiomyocytes or human primary cardiac fibroblasts where dysregulation of the given gene was statistically significant across all four conditions. Statistical significance of differential expression: p<0.05, Wald test with BH correction, n=4. Statistical significance of overlap of shared dysregulation between the four conditions: p<0.05, one-sided binomial test. RASG.=RASGEF1B. ( L ) Beta-gal staining quantifying senescence of primary human cardiac fibroblasts in response to predicted pro-aging perturbations (AAV overexpression of indicated genes). *p<0.05, Wilcoxon rank sums with BH correction, empty n=6, P4HA1 and RASGEF1B n=4. ( M ) Schematic of in vivo validation experiment. o/e=overexpression. ( N ) In vivo systolic function echocardiographic measurements at 6 weeks post-AAV9 injection. EF: GFP n=8, P4ha1 n=10, Rasgef1b n=10; GLS: GFP n=8, P4ha1 n=9, Rasgef1b n=8. *p<0.05 one-way ANOVA with multiple hypothesis correction.

Article Snippet: Human cardiac capillary endothelial cells were purchased from PromoCell (C-12285) and cultured in endothelial basal medium (EBM, CC-3121, Lonza) supplemented with 10% fetal bovine serum (10100147, Gibco).

Techniques: In Silico, Inhibition, Expressing, Staining, Gene Expression, RNA Sequencing, Over Expression, Derivative Assay, Quantitative Proteomics, In Vivo, Biomarker Discovery, Injection