Matrisome ratios turn fibroblast stress into a measurable phenotype space
The extracellular matrix is no longer just a scaffold. In fibroblasts, it acts as a durable biological record of stress, repair, and aging, preserving signals that outlast the molecular noise of a single transcriptome snapshot. ECMExplainer translates that record into a functional matrisome framework built from 180 gene-pair ratios, each ratio comparing whether one matrisome gene exceeds another in expression rather than depending on absolute counts. The approach is designed to classify fibroblast transcriptomes across six reference states: Proliferating, Quiescence, Senescence, Fibrosis, Inflammation, and Hypoxia.
This is a clever computational choice with practical biological consequences. Ratio-based features are far less vulnerable to sequencing depth shifts and batch offsets than magnitude-based classifiers, which means a sample can be interpreted even when the raw expression scale is unstable. Instead of forcing a transcriptome into a single label, ECMExplainer assigns coordinate distances to each reference phenotype, allowing a sample to sit near one state, drift between several, or remain outside the known training territory while still gaining biological meaning.
In fibroblast biology, that distinction matters. Stress responses rarely switch on in neat bins.
A compact ratio system with direct relevance to aging biology
The framework divides its 180 matrisome pairs into 76 Assembly/Destruction pairs and 104 Signaling/Signaling pairs, a structure that mirrors the two broad tasks of extracellular matrix biology: building and remodeling the matrix itself, then relaying the signals that tell cells how to behave within it. That balance is central to fibroblast identity. A cell that is actively proliferating, senescent, fibrotic, inflamed, or hypoxic does not merely change one gene at a time; it rewires the matrix program that defines tissue mechanics and intercellular communication.
Across 20 independent fibroblast datasets, ECMExplainer identified conserved ECM signatures spanning diverse stress-associated states. The study’s central claim is not that the model produces a prettier heat map than older classifiers, but that it captures a functional phenotype space that remains readable across datasets and conditions. Because the ratios are binary in spirit — one gene higher than another — the framework avoids some of the fragility that plagues magnitude-based methods when transcriptomes are affected by platform differences or technical offsets.
The result is a model that does more than label. It measures trajectory.
Senescence, fibrosis, and rejuvenation diverge along the ECM axis
One of the clearest examples involved FST/ITGA1, which emerged as a rapamycin-sensitive senescence-associated ratio. Rapamycin is biologically plausible in this setting because mTOR-linked signaling influences growth arrest, proteostasis, and stress adaptation, all of which can reshape how fibroblasts remodel their surrounding matrix. A ratio that shifts with rapamycin exposure suggests that the ECM state is not merely a downstream bystander of senescence but part of the senescence program itself.
The model also separated Hutchinson-Gilford progeria syndrome from chronological aging. In that context, ECMExplainer detected a fibrosis-biased ECM trajectory, implying that premature aging is not just accelerated time but a qualitatively different route through matrix remodeling. Fibrosis is not a trivial side branch of aging biology; it can dominate tissue behavior by stiffening the extracellular environment, amplifying mechanotransduction, and locking cells into persistent maladaptive states. A framework that recognizes that divergence is useful because two aging processes may look superficially similar in bulk transcriptomics while occupying very different ECM coordinates.
OSKM reprogramming left a measurable ECM aging footprint
The most operationally useful demonstration may be the ability to distinguish successful from failed OSKM reprogramming using an ECM Aging Score without retraining the classifier. OSKM reprogramming aims to reset cellular identity, but the extracellular matrix often remembers prior stress long after intracellular markers begin to shift. An ECM-based score is therefore attractive because it captures what the cell has done to its environment, not only what its nucleus currently says.
That point carries direct biological logic. Reprogramming can alter transcriptional state without fully erasing matrix history, especially in fibroblasts, where matrix deposition and remodeling can become self-reinforcing. A cell may appear partially reprogrammed on one axis yet retain a fibrotic or senescent ECM signature on another. ECMExplainer appears to read that mismatch rather than smoothing it away.
Why ratio-based matrisome modeling may travel well across datasets
The broader appeal of the framework lies in interpretability. Deep learning systems can classify cells well, but they often leave clinicians and biologists with opaque feature space. ECMExplainer keeps the biology visible by anchoring interpretation to named gene pairs and reference states. That makes the output easier to inspect when a sample falls between canonical phenotypes or when a tissue state reflects overlapping stress programs such as inflammation layered on top of hypoxia.
This design also raises an important translational possibility for aging research. If matrisome ratios truly preserve state information robustly across platforms, they could become a compact readout for tissue stress, matrix remodeling, and rejuvenation experiments. Fibroblasts are especially relevant because they sit at the center of wound repair, scarring, stromal support, and age-related tissue stiffening. A matrix-aware classifier may therefore be more informative than a single-gene aging marker when the biological question concerns tissue behavior rather than isolated molecular changes.
The preprint does not claim clinical readiness, and it should not be treated as a diagnostic tool. Its strength lies in conceptual coherence and in the way it connects extracellular matrix biology to aging trajectories, senescence, fibrosis, inflammation, hypoxia, and reprogramming. The next test is whether the same ratio framework generalizes beyond fibroblasts and whether specific matrisome coordinates predict functional tissue outcomes rather than only transcriptomic similarity.