We hypothesize that the exponential coefficient observed in aging pseudotime trajectories directly reflects the rate of mitochondrial NAD+ turnover in a given cell type, and that alterations in this metabolic rate shift the position of cells along the quiescence‑senescence continuum. In other words, tissues with higher NAD+ consumption exhibit faster exponential aging pseudotime, pushing cells toward senescent states, whereas tissues that maintain NAD+ pools show slower exponential dynamics and prolonged quiescence.

Mechanistic basis

• Pseudotime models (Monocle, DPT, Palantir) capture a latent progress variable that correlates with mortality risk when fitted with two exponentials 3.
• Mitochondrial NAD+ is a key regulator of sirtuin activity, DNA repair, and metabolic signaling, all of which influence epigenetic drift and transcriptional aging signatures.
• Single‑cell multi‑omics studies have shown concurrent variation in NAD+ metabolic gene expression and epigenetic age across cells 4.
• We propose that the exponential term’s rate constant (k) is proportional to the average mitochondrial NAD+ flux per cell, making pseudotime a readout of metabolic aging speed.

Testable predictions

1. Perturbing mitochondrial NAD+ biosynthesis (e.g., by overexpressing or inhibiting NAMPT) will shift the exponential coefficient k of pseudotime trajectories in a predictable direction: increased NAD+ flux → larger k (faster aging); decreased NAD+ flux → smaller k (slower aging).
2. The shift in k will be mirrored by a corresponding movement of cells along the quiescence‑senescence continuum, measurable by changes in the expression of canonical quiescence markers (e.g., FOXO3, p27) and senescence markers (e.g., p16, SASP factors).
3. Tissue‑specific differences in baseline k observed across blood, brain, and muscle will correlate with independently measured mitochondrial NAD+ turnover rates obtained via isotopic tracing or NAD+ biosensors.

Experimental design

• Step 1: Generate single‑cell RNA‑seq + ATAC‑seq datasets from primary human fibroblasts, hepatocytes, and neurons under three conditions: (a) control, (b) NAMPT overexpression, (c) NAMPT knockdown (using CRISPRa/i or small‑molecule modulators).
• Step 2: Compute pseudotime using multiple algorithms (Monocle, DPT, Palantir, psupertime) to ensure robustness against method selection bias 5.
• Step 3: Fit each trajectory with a double‑exponential model and extract the fast‑phase rate constant k_fast.
• Step 4: Quantify mitochondrial NAD+ flux in parallel using a genetically encoded NAD+ sensor (SoNar) or mass‑isotopic tracing of labeled nicotinamide.
• Step 5: Correlate k_fast with NAD+ flux across conditions and cell types; assess marker expression shifts along the pseudotime axis to locate quiescence‑senescence transition points.

Potential outcomes and falsification

• If NAD+ flux positively predicts k_fast and the transition point moves toward senescence with higher flux, the hypothesis is supported.
• If altering NAD+ levels fails to change the exponential coefficient, or if changes in k_fast occur without corresponding shifts in quiescence‑senescence markers, the hypothesis is falsified.
• Additionally, if method‑dependent variations in pseudotime outweigh the metabolic signal, we would need to refine the hypothesis to include algorithm‑specific correction factors.

This framework links a concrete metabolic mechanism to the abstract exponential dynamics of aging pseudotime, offers a clear route for experimental validation, and distinguishes true biological signal from analytical artefacts.