Conditioning clocks on organ function: what proteomic and epigenetic ageing measures actually do about filtration

Question: How do proteomic and epigenetic biological ageing clocks handle confounding by the physiological function of the tissue or biofluid they are measured in — such as kidney filtration for urine-derived markers?

Answer

Across the retrieved literature, the field has exactly one explicit, named solution to this problem and one implicit, general one — and neither is routine. The explicit solution is two-stage post hoc calibration on the functional readout: an unconditioned urinary peptidomic clock was built in a kidney-function-preserved derivation cohort and then conditioned on eGFR and urinary albumin-to-creatinine ratio by "Filtrate-Aware Calibration" fitted in an independent kidney-diverse cohort ([12]). The implicit solution is cell-type deconvolution, which is how DNA methylation clocks absorb the compositional consequences of tissue physiology rather than the physiology itself ([21], [3]). Elsewhere, confounding is simply left in: plasma proteomic clocks absorb impaired kidney function and specific drug-targeted proteins into age acceleration without any adjustment ([11], [10]), and organ-specific plasma clocks are the least accurate members of their class ([29]). This survey's position is that conditioning on organ function should be the default for analyte-abundance clocks measured in a clearance organ or fluid, but it is only defensible when accompanied by evidence that the conditioning removed handling rather than ageing — because in the kidney the functional lesion and the epigenetic ageing signature are close to the same measurement ([1], [30]).

Only one clock treats filtration as a named confounder, and it solves it by calibration rather than by design

The urinary peptidomic clock is the sole retrieved study that states the problem explicitly: an unconditioned clock (UPBioAge) was developed in a kidney function-preserved cohort of 1,811, then re-fitted on eGFR and UACR in an independent kidney-diverse cohort of 7,798 to produce k-UPBioAge ([12]). Calibration preserved accuracy — a holdout MAE of 4.91 years and r = 0.945, replicated at 5.43–5.47 years in two external cohorts — and kidney-conditioned age acceleration still predicted all-cause mortality (HR 1.48 per SD), incident coronary artery disease, heart failure and CKD progression ([12]). The one place the conditioned clock failed was the extreme stratum: no association was evident in participants with eGFR 15–29 mL/min/1.73 m², which the authors report as a boundary rather than explain ([12]). That boundary is the empirical signature of the trade-off this survey flags: heavy conditioning restores transportability across kidney function but strips out the participants in whom renal biology is most abnormal.

Urinary and plasma protein markers are substantially readouts of renal handling, so unadjusted clocks are partly filtration assays

Genetic evidence establishes that urine protein levels are under the control of the reabsorption machinery itself: variants in CUBN and LRP2, encoding the receptor complex for tubular protein reuptake from ultrafiltrate, were each associated with more than 20 urine proteins, and 38 of 174 loci showed urine-specific rather than plasma-shared signals ([18]). Detectability in urine is bimodal, with 47% of 2,868 assayed proteins below 20% detectability and highly detectable proteins enriched for proximal tubule origin, and only 157 proteins associated with age ([16]). Yet the confounding is not total, and the same literature shows urine can be a privileged compartment: vesicular albumin and podocalyxin were only moderately correlated with urinary albumin and not with eGFR ([14]), and a urinary extracellular-vesicle MUC1 assay is explicitly framed as bypassing the "systemic metabolic noise" that confounds blood markers ([23]). In plasma, the confound is equally visible: within-individual transitions to higher proteomic age acceleration tracked impaired kidney function and dyslipidaemia ([11]), CST3 — the canonical filtration marker — was among the three strongest mortality-associated plasma proteins ([17]), and medication initiation shifted clocks through single drug-targeted proteins such as renin and APOB rather than through generalised organ ageing ([10]).

Cell-type deconvolution is the epigenetic analogue of functional conditioning, and it only partly works

The methylation field has long known that bulk clock accuracy is partly a compositional artefact: approximately 39% of a blood clock's accuracy is driven by shifts in lymphocyte subsets and about 12% by neuronal subsets in brain, and cell-type-specific clocks detect disease acceleration that non-specific clocks miss ([21]). DNA methylation age is strongly associated with naive and memory T cell proportions, although age acceleration is far less so and, for second-generation clocks, mainly with neutrophils ([3]). In saliva the equivalent effect is buccal-versus-immune proportion: age acceleration differed significantly across stratified subsamples when cell-type proportions were not accounted for ([25]). But adjustment does not restore interchangeability. In matched pediatric blood and saliva, age acceleration estimates showed only low-to-moderate cross-tissue correlations (r = 0.20–0.68) and generalisability was limited "even when accounting for cell type composition" ([24]); saliva returned methylation ages 3.83–16.46 years higher than buffy coat ([26]); and CpG sets predicting the same phenotypes from cheek mucosa versus blood showed minimal overlap ([27]). Deconvolution removes the compositional channel but leaves the tissue-intrinsic channel untouched.

Building specificity in beats adjusting it out

Where tissue-matched models exist, they outperform adjusted pan-tissue clocks: a breast-specific clock reached r = 0.88 and MAE 3.27 years without dataset-specific regression adjustment ([22]), a clock optimised for menstrual fluid and endometrium outperformed existing clocks in a hormone-dependent tissue where they perform poorly ([4]), and pseudobulk single-nucleus ATAC clocks resolved asynchronous cell-type ageing invisible to bulk chromatin clocks ([19]). The strongest existence proof that composition sensitivity is a design choice rather than a necessity comes from a human age model built from penguin orthologues of ageing genes, which predicted age accurately while remaining largely insensitive to blood cell composition ([20]). On the proteomic side, the analogous move is being attempted — organ-enriched marker panels were rebuilt from organ-specific proteome data precisely because bulk transcriptomic annotation "may lead to the potential misidentification of protein markers" ([15]) — but organ specificity is acknowledged to shape generalisability and interpretability without any calibration on organ function ([28]).

Conditioning risks deleting the signal, because in the kidney disease state and ageing state converge

The decisive objection to naive eGFR conditioning is that renal pathology produces an epigenetic signature that mirrors ageing: in a cross-species single-cell kidney atlas, accelerated epigenetic ageing was dominated by tubular epithelia in diseased kidneys, and the pathological state "mirrors transcriptional trajectories observed in normal aging" ([1]). Blood methylation corroborates the entanglement — CpGs associated with late diabetic kidney disease progression were strongly associated with baseline eGFR, and the lead CpG cg17944885 is described as "a frequent finding in eGFR EWASs" ([30]). Reference-tissue work shows the same overlap from the other direction: biological processes associated with ageing were enriched in older kidney reference tissues, potentially overlapping with disease mechanisms, and adjusting for donor age changed pathway conclusions ([13]). Conditioning a clock on eGFR therefore removes, by construction, variance that includes genuine renal ageing.

Whether physiological state is a confounder depends on the training target, not on the biofluid

Clocks differ in perturbation sensitivity as a consequence of their construction rather than as a defect: an overnight fast with refeeding shifted immune-sensitive clocks by 0.5–3 years while within-person reliability remained high (ICC median ~0.96) ([2]). Biological reliability is markedly lower than technical reliability across 18 clocks, and technical robustness does not predict biological stability ([5]). Correspondingly, a deep-learning clock trained on clustered CpG sites reported that sex, BMI, smoking and other biological-age measures did not affect chronological age inference ([8]), whereas mortality-trained and organ-specific clocks are exactly where accuracy is weakest — low-accuracy clocks, "including many organ-specific clocks," provided little evidence for confidently accelerated or decelerated ageing ([29]). Chronological-age kernels appear relatively robust to physiological state; phenotypic and mortality kernels are where the confounding concentrates.

Conflicts and gaps

The literature disagrees on whether urine should be conditioned at all: the filtrate-aware clock assumes urinary peptides largely index filtration and corrects accordingly ([12]), while vesicle-based work reports renal-function-independent urinary signals and treats urine as a way of escaping systemic noise ([14], [23]). No retrieved plasma proteomic clock reports a pre-specified, eGFR-conditioned variant alongside its unconditioned version, so the magnitude of the plasma-side artefact is unquantified even though impaired kidney function is known to track acceleration ([11]). No study tests whether conditioning removes genuine renal ageing, despite the tubular-epithelial evidence that it would ([1]). No urinary or cell-free DNA methylation clock is conditioned on filtration at all; the cfDNA clock literature addresses sequencing depth and platform noise and is silent on tissue-of-origin mixture ([7]), while blood-trained GrimAge has been applied directly to follicular fluid cells with adjustment only for age and technical covariates ([6]). Finally, genetic confounding is a parallel and unaddressed axis: clock CpGs are heavily enriched for blood meQTLs (62.7% versus 39.6% for unused probes), so some between-person clock variance reflects inherited variation rather than ageing ([9]).

What would change the answer

Three concrete designs would settle it. First, a four-arm head-to-head in one kidney-diverse urine cohort — unconditioned, Filtrate-Aware-Calibrated, eGFR-stratified validation, and Mendelian randomisation using CUBN/LRP2 variants as instruments for tubular reabsorption ([18]) — would separate handling-driven from ageing-driven variance without statistical deletion. Second, transplanting the FAC procedure into a large plasma proteomic resource and reporting mortality hazard ratios for kidney-age acceleration before and after eGFR/UACR conditioning would quantify the plasma-side artefact that is currently only inferable ([12], [11]). Third, a within-person natural experiment exploiting drug-induced filtration change — the medication-initiation design already used to show that clocks move through single drug-targeted proteins rather than organ ageing ([10]) — would establish causal direction, and pairing it with uncertainty-aware reporting of prediction intervals ([29]) would show whether conditioning-invariant estimates remain informative. Applying single-cell kidney methylome references ([1]) to urinary cell pellets would additionally test whether urine-derived methylation age measures the tubular epithelium or the organism.

References

[1] Jeong, H., Lake, B. B., Diep, D., Li, X., Yan, Q., Gisch, D. L., Reinert, S., Eadon, M. T., Gaut, J. P., Jain, S., Zhang, K.. "A cross-species single-cell epigenome kidney atlas identifies epithelial cells as a driver of epigenetic aging". bioRxiv. 2026-01-23T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.22.700871

[2] Seale, K. B., Dwaraka, V. B., Giosan, I., Mendez, T., Smith, R.. "Fasting Status and Epigenetic Clock Stability: Implications for Aging Research". bioRxiv. 2026-06-07T00:00:00.000Z. URL: https://doi.org/10.64898/2026.06.04.729731

[3] Jonkman, T. H., van Zwet, E. W., Heijmans, B. T.. "Probing epigenetic clocks as a rational markers of biological age using blood cell counts". medRxiv. 2025-05-13T00:00:00.000Z. URL: https://doi.org/10.1101/2025.05.12.25327213

[4] Pavuluri, A., Gould, B., Indap, A., Salakh, N., Lacob, K., Dantas, A., Sazonova, O., Ching, J.. "The Xella Clock: a female-specific epigenetic aging clock optimized for menstrual fluid and endometrial tissue". medRxiv. 2026-08-10T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.07.26359971

[5] Sehgal, R., Borrus, D. S., Gonzalez, J. T., Markov, Y., ADNI,, Higgins-Chen, A. T.. "Biological versus Technical Reliability of Epigenetic Clocks and Implications for Disease Prognosis and Intervention Response". bioRxiv. 2025-10-14T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.13.682176

[6] Young, A. S., Campbell, K. A., Everson, T. S., Gennings, C., Braselton, M. E., Mullins, C. E., Jariwala, P., Smith, A. K., Spencer, J. B., Hipp, H., Gaskins, A. J., Walker, D. I.. "Interaction between the Chemical Exposome and Epigenome in Follicular Fluid from Patients Undergoing Assisted Reproduction". medRxiv. 2026-08-25T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.21.26361034

[7] Li, G., Huang, W., Zhao, X., Wu, J., Guo, Y., Chen, L., Cao, X., Yang, Z., Jiang, S., Hu, B., Wang, Y., Tan, D., Tong, V., Tang, C., Feng, X., Hu, X., Ouyang, C., Zhou, G.. "A Robust and Integrated Framework for Cross-platform Adaptation of Epigenetic Clocks in Cell-free DNA Sequencing". bioRxiv. 2026-06-03T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.27.690895

[8] Ochana, B.-L., Nudelman, D., Cohen, D., Peretz, A., Piyanzin, S., Gal, O., Horn, A., Loyfer, N., Varshavsky, M., Raisch, R., Shapiro, I., Friedlander, Y., Glaser, B., Hochner, H., Dor, Y., Kaplan, T., Shemer, R.. "Time is encoded by methylation changes at clustered CpG sites". bioRxiv. 2024-12-05T00:00:00.000Z. URL: https://doi.org/10.1101/2024.12.03.626674

[9] Sean Lim. "GIA: Germline-Informed Aging with AlphaGenome Finds Genetically Regulated CpGs". arxiv. 2026-09-17T00:00:00.000Z. URL: https://arxiv.org/abs/2609.17801

[10] Neirynck, R. E., Chirinos, J. A., Van Damme, M., Coussement, L., Segers, P., De Buyzere, M., Rietzschel, E. R., De Meyer, T.. "Longitudinal dynamics of organ-specific proteomic aging clocks over a decade of midlife". bioRxiv. 2026-02-18T00:00:00.000Z. URL: https://doi.org/10.64898/2026.02.17.706320

[11] Lermer, R., Qian, X., Pusch, S. T., Schaefer, T., Adam, J., Li, L., von Scheidt, M., Linkohr, B., Peters, A., Mann, M., Schunkert, H., Heinig, M.. "Intraindividual longitudinal changes of the proteome reveal dynamic markers of biological aging". bioRxiv. 2026-10-06T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.01.755909

[12] Biglari, S., Jaimes-Campos, M. A., Siwy, J., Latosinska, A., Mischak, H., Nawrot, T. S., Staessen, J. A., Martens, D. S., Banasik, M.. "A kidney-conditioned urinary peptidomic biological ageing clock predicts all-cause mortality and age-related health outcomes". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.09.01.26361640

[13] Menon, R., Kimmel, P. L., Otto, E. A., Subramanian, L., OConnor, C. L., Godfrey, B., Smith, C., Alakwaa, F. M., Berthier, C. C., Sarwal, M., Woodle, E. S., Pyle, L., Choi, Y. J., Ladd, P., Sedor, J. R., Rosas, S. E., Waikar, S. S., Naik, A. S., Melo Ferreira, R., Eadon, M. T., Bitzer, M., Bjornstad, P., Hodgin, J. B., Kretzler, M.. "Not all controls are made equal: Definition of human kidney reference samples by single cell gene expression profiles". medRxiv. 2026-01-16T00:00:00.000Z. URL: https://doi.org/10.1101/2025.03.17.25324134

[14] Schnobrich, L., de Hesselle, H. C., Mornhinweg, L., Felgner, R., Zimmermann, M. E., Brandl, C., Heid, I. M., Castrop, H., Stark, K. J.. "Urinary vesicle biomarkers and kidney function: Results from the German AugUR study". medRxiv. 2026-01-13T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.12.26343937

[15] Xu, H., Chen, J., Chen, D., Mao, K., HAN, J.-D. J.. "Proteome-aware organ proxy aging clocks". bioRxiv. 2026-04-28T00:00:00.000Z. URL: https://doi.org/10.64898/2026.04.24.720503

[16] Meier, M., Schoberth, M., Li, Y., Borisov, O., Uluvar, B., Willecke, C., Meiselbach, H., Eckardt, K.-U., Schilling, O., Kottgen, A., Haug, S.. "Characterization of the Urine Proteome Using the Olink Explore 3072 Platform". medRxiv. 2025-12-09T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.08.25341840

[17] Pyrgioti, M., Eguiagaray, I. M., Redmond, P., Corley, J., Bastin, M. E., Valdes Hernandez, M., Russ, T. C., Wardlaw, J. M., Hannon, E., Deary, I. J., Walker, K. A., Tucker-Drob, E. M., Cox, S. R., Marioni, R. E., Harris, S. E.. "Multimodal Ageing Biomarkers and Plasma Proteomic Signatures Associated with All-Cause Mortality". medRxiv. 2026-03-10T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.09.26347914

[18] Haug, S., Borisov, O., Lehner, N., Schoberth, M., Menneking, I., Scherer, N., Xu, K., Khan, A., Monteiro-Martins, S., Uluvar, B., Roemisch-Margl, W., Arnold, M., Schlosser, P., Sekula, P., Kronenberg, F., Eckardt, K.-U., Wuttke, M., Kastenmueller, G., Li, Y., Kottgen, A.. "Insights into Renal Protein Handling Through GWAS of the Human Urine Proteome". medRxiv. 2025-12-09T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.08.25341835

[19] Yu, P., Yu, D., Xue, Y., Noble, W. S.. "Cell type-resolved chromatin accessibility clocks for brain aging". bioRxiv. 2026-08-25T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.22.746439

[20] Hukkanen, M., Nitta Fernandes, F. A., Davis, L. R., Hofer, M., Bardon, G., Paciello, E., Zahn, S., Zhu, T., Westerlund, S., Meyer, B. S., Ollikainen, M., Le Bohec, C., Cristofari, R.. "An epigenetic palimpsest: the King penguin's blood methylome reveals conserved hematopoietic ageing". bioRxiv. 2026-10-13T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.10.757848

[21] Tong, H., Guo, X., Jacques, M., Luo, Q., EYNON, N., Teschendorff, A.. "Cell-type specific epigenetic clocks to quantify biological age at cell-type resolution". bioRxiv. 2024-11-22T00:00:00.000Z. URL: https://doi.org/10.1101/2024.07.30.605833

[22] Garma, L. D., Pernas, S., Baz, D. V., Garcia Campelo, R., Terrasa, J., Colomer, R., Jimenez, D., Vera Garcia, R., Bermejo, B., Gonzalez Santiago, S., Quintela-Fandino, M.. "A breast tissue-specific epigenetic clock provides accurate chronological age predictions and reveals de-correlation of age and DNA methylation in tumor-adjacent and tumor samples". bioRxiv. 2025-02-28T00:00:00.000Z. URL: https://doi.org/10.1101/2025.02.21.639537

[23] Nishimura, T., Harita, Y., Hirakawa, Y., Takizawa, K., Fujishiro, J., Ogawa, S., Kajiho, Y., Kanda, S., Kushima, R., Omori, T., Hamasaki, Y., Gotoh, Y., Miura, K., Fujita, N., Okamoto, T., Hisano, M., Nangaku, M., Kato, M.. "A non-invasive liquid biopsy resolves the diagnostic blind spot in chronic kidney disease". medRxiv. 2026-06-17T00:00:00.000Z. URL: https://doi.org/10.64898/2026.06.09.26355142

[24] Karlbauer, V. N., Dominguez-Baleon, C., Rex-Haffner, M., Namendorf, T., Hoffmann, F., Klawitter, H., Entringer, S., Buss, C., Winter, S., Heim, C., Binder, E. B., Czamara, D.. "Limited Generalizability of Epigenetic Clocks and Scores Across Pediatric Tissues and Age Ranges". medRxiv. 2025-12-16T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.11.25340872

[25] Chan, M. H.-M., Meijer, M., Merrill, S. M., Fu, M. P. Y., Lin, D., MacIsaac, J. L., Riis, J. L., Granger, D. A., Thomas, E. A., Kobor, M. S.. "Not All Saliva Samples Are Equal: The Role of Cellular Heterogeneity in DNA methylation and Epigenetic Age Analyses with Biological and Psychosocial Factors". bioRxiv. 2025-05-08T00:00:00.000Z. URL: https://doi.org/10.1101/2024.11.08.621377

[26] Bruellman, R., Evans, D., Smolen, A., Evans, L. M., Reynolds, C. A.. "Differential DNA methylation clock ages across buffy coat (BC), peripheral blood mononuclear cells (PBMC), and saliva in individuals in early-to-mid adulthood". bioRxiv. 2026-03-24T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.16.673560

[27] Shoji, T., Tomo, Y., Nakaki, R.. "Predicting Biological Age and Clinical Biomarkers from DNA Methylation Profiles of Cheek Mucosa". bioRxiv. 2026-05-14T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.12.724485

[28] WEN, J.. "Refining the generation, interpretation, and application of multi-organ, multi-omics biological aging clocks". medRxiv. 2025-06-10T00:00:00.000Z. URL: https://doi.org/10.1101/2025.02.06.25321803

[29] Wang, C., Wu, H., Namba, S., Park, J. Y., Matsuda, K., Okada, Y., He, Z., Ionita-Laza, I.. "Beyond point estimates: quantifying predictive uncertainty reveals hidden dimensions of biological age acceleration and improves risk interpretation". bioRxiv. 2026-08-05T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.05.742349

[30] Syreeni, A., Dahlström, E. H., Smyth, L. J., Hill, C., Mutter, S., Gupta, Y., Harjutsalo, V., Chen, Z., Natarajan, R., Krolewski, A. S., Hirschhorn, J. N., Florez, J. C., GENIE Consortium,, Maxwell, A. P., Groop, P.-H., McKnight, A. J., Sandholm, N., FinnDiane Study Group,. "Blood methylation biomarkers are associated with diabetic kidney disease progression in type 1 diabetes". medRxiv. 2024-11-29T00:00:00.000Z. URL: https://doi.org/10.1101/2024.11.28.24318055

How this survey was made

This survey answers one question: “How do proteomic and epigenetic biological ageing clocks handle confounding by the physiological function of the tissue or biofluid they are measured in, such as kidney filtration for urine-derived markers?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 30 papers whose work this text draws on, chosen from the 113 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 21, 2026 and covered work from the previous two years.