Introduction to OmniAgeR

Zhaozhen Du1* and Andrew E. Teschendorff1**

1Shanghai Institute of Nutrition and Health, CAS

*duzhaozhen2022@sinh.ac.cn
**andrew@sinh.ac.cn

Package

OmniAgeR 0.99.4

Contents

1 Overview of Implemented Clocks

The OmniAgeR package provides two primary, high-level interfaces: epiMarker(): For calculating a comprehensive suite of aging-related clocks (chronological, biological, mitotic, etc.). For advanced users or specific applications, the package also provides direct access to individual clock functions (e.g., horvath2013Clock()) and specialized “bundle” calculators (e.g., pcClocks(), systemsAge()) that manage complex dependencies. These tools are broadly categorized into three main groups:

1. Epigenetic (DNAm) Clocks
2. Transcriptomic (RNA) Clocks
3. Other Aging-Related Surrogate Biomarkers

The suite of epigenetic aging clocks is particularly extensive, organized into several functional categories, including predictors for chronological age, biological age, cellular division (mitotic clocks), and gestational age.

2 Epigenetic (DNAm) Aging Clocks

2.1 Chronological Age Clocks

These DNAm clocks are designed for the prediction of chronological age.

• Horvath2013: (Horvath 2013) The original pan-tissue clock from 353 CpGs.
• Hannum: (Hannum et al. 2013) Blood-specific clock from 71 CpGs.
• Lin: (Lin et al. 2016) Clock based on 99 CpGs.
• VidalBralo: (Vidal-Bralo et al. 2016) Clock based on 8 CpGs.
• ZhangClock: (Zhang et al. 2019) Improved-precision clock from 514 CpGs.
• Horvath2018: (Horvath et al. 2018) Skin & blood clock.
• Bernabeu_cAge: (Bernabeu et al. 2022) Clock based on 3225 CpGs.
• CorticalClock: (Shireby et al. 2020) Brain cortical clock.
• PedBE: (McEwen et al. 2019) Pediatric buccal epithelial clock for children.
• CentenarianClock: (Eric Dec et al. 2023) Centenarian epigenetic clocks.
• Retro_age: (Ndhlovu et al. 2024) Retroelement-based clock developed on the EPIC v1/v2 arrays.
• ABEC: (Lee et al. 2020) Adult blood-based EPIC clock.
• eABEC: (Lee et al. 2020) Extended adult blood-based EPIC clock.
• cABEC: (Lee et al. 2020) Common adult blood-based EPIC clock.
• PipekElasticNet: (Pipek et al. 2023) Pipek’s multi-tissue elastic Net epigenetic clock (239 CpGs).
• PipekFilteredh: (Pipek et al. 2023) Pipek’s filtered horvath epigenetic clock (272 CpGs).
• PipekRetrainedh: (Pipek et al. 2023) Pipek’s retrained horvath epigenetic clock (308 CpGs).
• WuClock: (Wu et al. 2019) Wu’s Epigenetic Clock for Pediatric age estimation.
• Weidner: (Weidner et al. 2014) Calculate weidner epigenetic age (3 CpGs).
• IntrinClock: (Tomusiak et al. 2024) Calculates the intrinsic cellular age.
• Garagnani: (Garagnani et al. 2012) The Garagnani ELOVL2-based epigenetic age score(1 CpG).
• PCHorvath2013, PCHorvath2018, PCHannum: (Higgins-Chen et al. 2022) Computationally-bolstered versions of the original clocks. These PC clocks are retrained using principal components (PCs) derived from a large CpG set to minimize technical noise and improve reliability

2.2 Biological Age Clocks

These advanced clocks are designed to capture biological aging rather than chronological time, often showing stronger associations with health outcomes and mortality.

• Zhang10: (Zhang et al. 2017) 10-CpG clock associated with mortality.
• PhenoAge: (Levine et al. 2018) Predicts phenotypic age from 513 CpGs.
• DunedinPACE: (Belsky et al. 2022) Quantifies the pace of biological aging.
• GrimAge1: The original GrimAge clock (Lu et al. 2019).
• GrimAge2: (Lu et al. 2022) Updated composite biomarker of mortality risk.
• PCPhenoAge, PCGrimAge1: (Higgins-Chen et al. 2022) Computationally-bolstered PC versions of the PhenoAge and GrimAge1 clocks, retrained on principal components for enhanced reliability.
• DNAmFitAge: (McGreevy et al. 2023) Biological age indicator incorporating DNAmGrimAge and 3 DNAm-based physical fitness markers.
• IC_Clock: (Fuentealba et al. 2025) The intrinsic capacity (IC) clock based on 91 CpGs.
• SystemsAge: (Sehgal et al. 2025) The systems age and 11 system-specific scores.

2.3 Cellular Aging Clocks

This category groups biomarkers that measure two key forms of cellular aging: cell proliferation (mitotic clocks) and telomere attrition (DNAmTL).

2.3.1 Mitotic Clocks (Cellular Division)

A specialized set of clocks that measure cell division history or stem cell divisions.

• epiTOC1: (Yang et al. 2016) Average beta value of 385 promoter CpGs.
• epiTOC2: (Teschendorff et al. 2020) Estimates stem cell divisions using a dynamic model.
• epiTOC3: Estimates stem cell divisions based on unmethylated population doubling associated CpGs.
• stemTOCvitro: (Zhu et al. 2024) The 0.95 upper quantile of 629 stemTOCvitro CpGs (promoter CpGs unmethylated in fetal tissue that hypermethylate with population-doublings).
• stemTOC: (Zhu et al. 2024) The 0.95 upper quantile of 371 stemTOC CpGs, filtered for in-vivo hypermethylation with age.
• RepliTali: (Endicott et al. 2022) Based on 87 population doubling associated hypomethylated CpGs.
• HypoClock: (Teschendorff et al. 2020) Based on hypomethylation at 678 solo-WCGW sites.
• EpiCMIT_hyper: (Duran-Ferrer et al. 2020) Average beta value of 184 age-associated hypermethylated CpGs.
• EpiCMIT_hypo: (Duran-Ferrer et al. 2020) Average beta value of 1164 age-associated hypomethylated CpGs.

2.3.2 DNAm Telomere Length (TL) Clocks

• DNAmTL: (Lu AT et al. 2019) Calculates the Leukocyte telomere length.
• PCDNAmTL: (Lu AT et al. 2019; Higgins-Chen et al. 2022) A computationally-bolstered “PC” version of the original DNAmTL clock. This clock is retrained using principal components (PCs) derived from a large CpG set to minimize technical noise and improve reliability.

2.4 Causal Clocks

A new generation of clocks developed to reflect potential causal drivers of aging.

• CausalAge, DamAge, AdaptAge: (Ying et al. 2024) Three clocks derived from a causality-enriched model. They dissect aging into distinct components: ‘Causal’ (CausalAge), ‘Damage’ (DamAge), and Adaptation’ (AdaptAge).

2.5 Stochastic Clocks

• StocH, StocP, StocZ: (Tong et al. 2024) Stochastic analogues of the Horvath, PhenoAge, and Zhang clocks, trained on artificial cohorts to quantify the stochastic component of epigenetic aging.

2.6 Cell-Type Specific Clocks

These clocks are designed to measure aging in specific cell types or tissues, accounting for cellular heterogeneity.

• Neu-In, Glia-In: (Tong et al. 2024) Intrinsic clocks for neurons, and glia.
• Neu-Sin, Glia-Sin, Hep: (Tong et al. 2024) Semi-intrinsic clocks for neurons, glia, and hepatocytes

2.7 Gestational Age Clocks

This collection includes DNAm clocks designed to predict gestational age at birth.

• BohlinGA: (Bohlin et al. 2016) The Bohlin Gestational Age.
• EPICGA: (Haftorn et al. 2021) The EPIC Gestational Age.
• LeeGA: (Lee et al. 2019) A list (LeeControl, LeeRobust, LeeRefinedRobust).
• KnightGA: (Knight et al. 2016) The Knight Gestational Age.
• MayneGA: (Mayne et al. 2017) The Mayne Placental Gestational Age.

2.8 Cross-Species Support

• EnsembleAge: (Haghani et al. 2025) The multi-clock framework implementation. This includes mouse-specific estimators (EnsembleAgeMouse_Dynamic and EnsembleAgeMouse_Static) and a cross-species (EnsembleAgeMouse_HumanMouse) predictor applicable to both humans and mice

• UniversalPanMammalianClocks: (Lu et al. 2023) The three pan-tissue clocks (Clock 1, 2, 3) applicable across all mammalian tissues.

• PanMammalianBlood: (Lu et al. 2023) The two blood-specific pan-mammalian clocks (Clock 2, 3).

• PanMammalianSkin: (Lu et al. 2023) The two skin-specific pan-mammalian clocks (Clock 2, 3).

3 Transcriptomic Clocks

In addition to DNAm-based models, the package implements cutting-edge transcriptomic (RNA-seq) clocks:

• sc-ImmuAging: A single-cell, immune-cell-type-specific aging clock.
• Brain_CT_clock: A brain-cell-type-specific aging clock based on Single-Nuclei Transcriptomics.
• PASTA: A robust, multi-tissue transcriptomic clock based on a novel age-shift learning strategy. It utilizes rank-normalization to ensure broad applicability across heterogeneous datasets, including bulk RNA-seq, single-cell pseudobulks, and microarrays.

4 Surrogate Biomarkers & Scores

Besides, OmniAgeR includes functions to compute several other important surrogate markers associated with aging, lifestyle, and health status from DNAm data.

• CRP: Calculates a score for C-reactive protein (CRP) levels, a marker of inflammation.

• CHIP: Calculates scores for Clonal Hematopoiesis of Indeterminate Potential (CHIP).

• IL6: Computes a DNA methylation (DNAm) surrogate score for Interleukin-6 (IL-6) protein levels.

• EpiScores: Computes the 109 validated epigenetic scores (EpiScores) that serve as DNA methylation-based proxies for the levels of circulating plasma proteins as defined by Gadd et al. (2022).

• McCartneyTrait: Computes epigenetic surrogate scores for 10 different complex traits and biomarkers based on blood methylation models. Traits include: BMI, Smoking (pack-years), Alcohol (units/week), Education (years), Total_cholesterol, HDL_cholesterol, LDL_cholesterol, Total_HDL_ratio, WHR (Waist-to-Hip Ratio), and Body_fat_Perc.

5 Disease Risk Prediction

OmniAgeR provides a suite of tools for disease risk assessment, including a DNA methylation-based smoking index derived from 1,501 CpG sites, as well as specialized scores for cancer risks. These tools leverage robust epigenetic signatures to quantify environmental exposures and clinical outcomes.

• CompSmokeIndex: Computes a DNAm-based Smoking Index that correlates with cancer risk across various tissues.
• HepatoXuRisk: Calculate HepatoXu ctDNA Methylation Scores for Hepatocellular Carcinoma

6 DNA Methylation Cell-Type Fraction

The OmniAgeR package also provides a DNA Methylation Cell-Type Fraction (CTF) clock (implemented as dnamCTFClock), which was trained on a large-scale NSPT dataset comprising more than 3,500 blood samples. The clock incorporates 12 immune cell types, estimated using the EpiDISH algorithm, to model aging based on cell-type composition.

7 Available Clocks and Functions

7.1 Epigenetic (DNAm) Aging Clocks

Clock Name	Category	Calculator / Function
Horvath2013	Chronological	epiMarker(),horvath2013Clock()
Hannum	Chronological	epiMarker(),hannumClock()
Lin	Chronological	epiMarker(),linClock()
VidalBralo	Chronological	epiMarker(),vidalBraloClock()
ZhangClock	Chronological	epiMarker(),zhangClock()
Horvath2018	Chronological	epiMarker(),horvath2018Clock()
Bernabeu_cAge	Chronological	epiMarker(),bernabeuCAge()
CorticalClock	Chronological	epiMarker(),corticalClock()
PedBE	Chronological	epiMarker(),pedBEClock()
CentenarianClock	Chronological	epiMarker(),centenarianClock()
Retro_age	Chronological	epiMarker(),retroAge()
ABEC	Chronological	epiMarker(),leeABEC()
eABEC	Chronological	epiMarker(),leeExtendedABEC()
cABEC	Chronological	epiMarker(),leeCommonABEC()
PipekElasticNet	Chronological	epiMarker(),pipekElasticNet()
PipekFilteredh	Chronological	epiMarker(),pipekFilteredh()
PipekRetrainedh	Chronological	epiMarker(),pipekRetrainedh()
WuClock	Chronological	epiMarker(),wuClock()
Weidner	Chronological	epiMarker(),weidnerClock()
IntrinClock	Chronological	epiMarker(),intrinClock()
Garagnani	Chronological	epiMarker(),garagnaniClock()
Zhang10	Biological	epiMarker(),zhang10()
PhenoAge	Biological	epiMarker(),phenoAge()
DunedinPACE	Biological	epiMarker(),dunedinPACE()
GrimAge1	Biological	epiMarker(),grimAge1()
GrimAge2	Biological	epiMarker(),grimAge2()
DNAmFitAge	Biological	epiMarker(),dnamFitAge()
IC_Clock	Biological	epiMarker(),icClock()
SystemsAge	Biological	epiMarker(),systemsAge()
epiTOC1	Mitotic	epiMarker(),epiTOC1()
epiTOC2	Mitotic	epiMarker(),epiTOC2()
epiTOC3	Mitotic	epiMarker(),epiTOC3()
stemTOC	Mitotic	epiMarker(),stemTOC()
stemTOCvitro	Mitotic	epiMarker(),stemTOCvitro()
RepliTali	Mitotic	epiMarker(),repliTali()
HypoClock	Mitotic	epiMarker(),hypoClock()
EpiCMIT_hyper, EpiCMIT_hypo	Mitotic	epiMarker(),epiCMIT()
DNAmTL	Telomere length	epiMarker(),dnamTL()
PCHorvath2013,PCHorvath2018, PCHannum, PCPhenoAge, PCGrimAge1, PCDNAmTL	Chronological/Biological/Telomere length	epiMarker(),pcClocks()
CausalAge,DamAge,AdaptAge	Causal	epiMarker(),causalClock()
StocH,StocP,StocZ	Causal	epiMarker(),stochClocks()
Neu-In,Glia-In,Neu-Sin,Glia-Sin,Hep	Cell-Type Specific	epiMarker(),ctsClocks()
BohlinGA	Gestational	epiMarker(),bohlinGa()
EPICGA	Gestational	epiMarker(),epicGa()
LeeGA	Gestational	epiMarker(),LeeGa()
KnightGA	Gestational	epiMarker(),knightGa()
MayneGA	Gestational	epiMarker(),mayneGa()
EnsembleAge_HumanMouse	Mouse & Human	epiMarker(),ensembleAge()
EnsembleAge_Static	Mouse & Human	epiMarker(),ensembleAge()
EnsembleAge_Dynamic	Mouse & Human	epiMarker(),ensembleAge()
UniversalPanMammalianClocks	PanMammalian	epiMarker(),universalPanMammalianClocks()
PanMammalianBlood	PanMammalian	epiMarker(),panMammalianBlood()
PanMammalianSkin	PanMammalian	epiMarker(),panMammalianSkin()
DNAm_CTF_Clock	Cell-Type Fraction Clock	epiMarker(),dnamCTFClock()

7.2 Transcriptomic Clocks

Clock Name	Category	Calculator / Function
scImmuAging	Immune-cell-type-specific	scImmuAging()
Brain_CT_Clock	Brain-cell-type-specific	brainCtClock()
PASTA	Multi-tissue	pastaScores()

7.3 Surrogate Biomarkers & Scores

Clock Name	Target	Calculator / Function
CRP	CRP/intCRP score	epiMarker(), compCRP()
CHIP	CHIP score	epiMarker(), compCHIP()
IL6	IL6 score	epiMarker(), compIL6()
EpiScores	109 validated epigenetic scores	epiMarker(), compEpiScores()
CompSmokeIndex	Smoking Index	epiMarker(), compSmokeIndex()
HepatoXuRisk	Hepatocellular Carcinoma Risk	epiMarker(), hepatoXuRisk()
SmokeIndex	Smoking Index	epiMarker(), compSmokeIndex()
McCartney_Trait	Epigenetic surrogate scores for 10 different complex traits	epiMarker(), mcCartneyTrait()

8 Tutorial Example 1: Apply mitotic clocks on Lung pre-cancerous lesions

8.1 Loading the lung pre-cancerous lesions dataset

This is an Illumina 450k dataset encompassing 21 normal lung tissue samples and 35 lung carcinoma in situ (LCIS) samples, of which 22 progressed to an invasive lung cancer (LC). Here we explore the correlation between mitotic age and cancer state with linear model, treating N, LCIS, and LC as ordinal variable (1,2,3) and adjusting for age.

library(OmniAgeR)
library(ggplot2)
library(patchwork)
library(ggpubr)
lungInv <- loadOmniAgeRdata(
    "omniager_lung_inv",
    verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

lungInvM <- lungInv$bmiq_m
phenoDf <- lungInv$PhenoTypes

my_comparisons <- list(c("N\nN=21", "LCIS\nN=13"), c("LCIS\nN=13", "LCIS->LC\nN=22"))
table(phenoDf$Group)

## 
##     LCIS\nN=13 LCIS->LC\nN=22        N\nN=21 
##             13             22             21

## Check available epigenetic clocks
listEpiMarker()

## $mitotic
## [1] "epiTOC1"       "epiTOC2"       "epiTOC3"       "stemTOCvitro" 
## [5] "stemTOC"       "RepliTali"     "HypoClock"     "EpiCMIT_Hyper"
## [9] "EpiCMIT_Hypo" 
## 
## $dnamtl
## [1] "DNAmTL"   "PCDNAmTL"
## 
## $chronological
##  [1] "Horvath2013"      "Hannum"           "Lin"              "VidalBralo"      
##  [5] "ZhangClock"       "Horvath2018"      "Bernabeu_cAge"    "CorticalClock"   
##  [9] "PedBE"            "CentenarianClock" "Retro_age"        "ABEC"            
## [13] "eABEC"            "cABEC"            "PipekElasticNet"  "PipekFilteredh"  
## [17] "PipekRetrainedh"  "WuClock"          "Weidner"          "IntrinClock"     
## [21] "Garagnani"        "PCHorvath2013"    "PCHorvath2018"    "PCHannum"        
## 
## $biological
##  [1] "Zhang10"     "PhenoAge"    "DunedinPACE" "GrimAge1"    "GrimAge2"   
##  [6] "PCPhenoAge"  "PCGrimAge1"  "DNAmFitAge"  "IC_Clock"    "SystemsAge" 
## 
## $causal
## [1] "CausalAge" "DamAge"    "AdaptAge" 
## 
## $cellTypeSpecific
## [1] "Neu-In"   "Neu-Sin"  "Glia-In"  "Glia-Sin" "Hep"     
## 
## $stochastic
## [1] "StocH" "StocZ" "StocP"
## 
## $gestationalAge
## [1] "BohlinGA" "EPICGA"   "KnightGA" "LeeGA"    "MayneGA" 
## 
## $surrogateBiomarkers
## [1] "CRP"       "CHIP"      "IL6"       "EpiScores"
## 
## $traitPred
## [1] "McCartneyTrait"
## 
## $diseaseRisk
## [1] "SmokeIndex"   "HepatoXuRisk"
## 
## $crossSpecies
## [1] "EnsembleAge_HumanMouse"      "EnsembleAge_Static"         
## [3] "EnsembleAge_Dynamic"         "UniversalPanMammalianClocks"
## [5] "PanMammalianBlood"           "PanMammalianSkin"

8.2 Estimation of epigenetic mitotic age

Mitotic age estimation of all the clocks is implemented in a general function EpiMitClocks. The required input is a DNAm beta value matrix with rows labeling Illumina 450k/EPIC CpGs and columns labeling samples.

epiMarkerRes <- epiMarker(
    betaM = lungInvM,
    clockNames = "mitotic",
    chronAge = phenoDf$Age,
    minCoverage = 0,
    verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

rm(lungInvM)

phenoDf$epiTOC1 <- epiMarkerRes$epiTOC1
phenoDf$epiTOC2 <- epiMarkerRes$epiTOC2$irS
phenoDf$epiTOC3 <- epiMarkerRes$epiTOC3$irS
phenoDf$stemTOCvitro <- epiMarkerRes$stemTOCvitro
phenoDf$stemTOC <- epiMarkerRes$stemTOC
phenoDf$HypoClock <- epiMarkerRes$HypoClock
phenoDf$RepliTali <- epiMarkerRes$RepliTali
phenoDf$EpiCMIT_Hyper <- epiMarkerRes$EpiCMIT_Hyper
phenoDf$EpiCMIT_Hypo <- epiMarkerRes$EpiCMIT_Hypo
g <- list()
for (i in 1:9) {
    tempDf <- data.frame("Group" = phenoDf$Group, "Age" = phenoDf$Age, "num" = phenoDf$num, "score" = phenoDf[, i + 3])
    y <- summary(lm(tempDf$score ~ tempDf$num + tempDf$Age))$coefficients[2, 4]
    g[[i]] <- ggplot(tempDf, aes(x = Group, y = score, fill = Group)) +
        guides(fill = "none") +
        geom_boxplot() +
        theme_bw() +
        scale_x_discrete(limits = c("N\nN=21", "LCIS\nN=13", "LCIS->LC\nN=22")) +
        stat_compare_means(method = "wilcox.test", comparisons = my_comparisons, size = 4) +
        xlab("") +
        ylab(colnames(phenoDf)[i + 3]) +
        annotate("text", x = 2, y = max(tempDf$score) * 0.9, label = paste0("P(Age-adjusted)=", signif(y, 2)), size = 4) +
        theme(
            axis.title = element_text(size = 11),
            axis.text = element_text(size = 11)
        )
}

ggpubr::ggarrange(plotlist = g, nrow = 3, ncol = 3)

9 Tutorial Example 2: Predict chronological age in blood tissue

The epigenetic age of 50 normal blood samples was estimated from their Illumina 450k methylation profiles using both the Horvath and Zhang epigenetic clocks.

library(OmniAgeR)
library(ggplot2)
library(patchwork)
library(ggpubr)
hannumExample <- loadOmniAgeRdata(
    "omniager_hannum_example",
    verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

hannumBmiqM <- hannumExample[[1]]
phenoTypesHannum <- hannumExample[[2]]
age <- phenoTypesHannum$Age
sex <- ifelse(phenoTypesHannum$Sex == "F", "Female", "Male")
epiMarkerOut <- epiMarker(hannumBmiqM,
    clockNames = c("Horvath2013", "ZhangClock"),
    minCoverage = 0, verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation
## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

gp <- list()

for (i in seq_along(epiMarkerOut)) {
    plot_df <- data.frame(ActualAge = phenoTypesHannum$Age, PredictedAge = epiMarkerOut[[i]])
    cor_test_result <- cor.test(plot_df$ActualAge, plot_df$PredictedAge)
    correlation <- cor_test_result$estimate
    p_value <- cor_test_result$p.value
    mae <- mean(abs(plot_df$PredictedAge - plot_df$ActualAge))
    p_value_formatted <- ifelse(p_value < 0.001,
        formatC(p_value, format = "e", digits = 2),
        round(p_value, 3)
    )
    annotation_text <- paste0(
        "R = ", round(correlation, 3), "\n",
        "P = ", p_value_formatted, "\n",
        "MAE = ", round(mae, 2)
    )
    gp[[i]] <- ggplot(data = plot_df, aes(x = ActualAge, y = PredictedAge)) +
        geom_point(color = "steelblue", size = 3, alpha = 0.8) +
        geom_smooth(method = "lm", color = "black", se = FALSE) +
        annotate("text",
            x = min(plot_df$ActualAge, na.rm = TRUE),
            y = max(plot_df$PredictedAge, na.rm = TRUE),
            label = annotation_text,
            hjust = 0,
            vjust = 1, size = 5
        ) +
        labs(
            title = NULL,
            x = "Chronological Age",
            y = paste0("Predicted Age(", c("Horvath2013", "ZhangClock")[i], ")")
        ) +
        theme_bw() +
        theme(
            plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
            plot.subtitle = element_text(hjust = 0.5, size = 12),
            axis.title = element_text(size = 14),
            axis.text = element_text(size = 12)
        )
}

ggpubr::ggarrange(plotlist = gp, nrow = 1, ncol = 2)

## `geom_smooth()` using formula = 'y ~ x'
## `geom_smooth()` using formula = 'y ~ x'

10 Tutorial Example 3: Predict gestational age

A small demonstration dataset containing three samples, used to illustrate how to predict the gestational age of samples using DNAm Gestational Age Clocks.

library(OmniAgeR)
library(ggplot2)
library(patchwork)
library(ggpubr)
gaExample <- loadOmniAgeRdata(
    "omniager_ga_example",
    verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

gaM <- gaExample[[1]]
phenoTypesGa <- gaExample[[2]]
epiGA <- epiMarker(gaM,
    clockNames = c("KnightGA", "MayneGA"),
    minCoverage = 0, verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation
## loading from cache

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

gp <- list()

for (i in seq_along(epiGA)) {
    plot_df <- data.frame(Gestational_Age = phenoTypesGa$age, PredictedAge = epiGA[[i]])
    cor_test_result <- cor.test(plot_df$Gestational_Age, plot_df$PredictedAge)
    correlation <- cor_test_result$estimate
    p_value <- cor_test_result$p.value
    mae <- mean(abs(plot_df$PredictedAge - plot_df$Gestational_Age))
    p_value_formatted <- ifelse(p_value < 0.001,
        formatC(p_value, format = "e", digits = 2),
        round(p_value, 3)
    )
    annotation_text <- paste0(
        "R = ", round(correlation, 3), "\n",
        "P = ", p_value_formatted, "\n",
        "MAE = ", round(mae, 2)
    )
    gp[[i]] <- ggplot(data = plot_df, aes(x = Gestational_Age, y = PredictedAge)) +
        geom_point(color = "steelblue", size = 3, alpha = 0.8) +
        geom_smooth(method = "lm", color = "black", se = FALSE) +
        annotate("text",
            x = min(plot_df$Gestational_Age, na.rm = TRUE),
            y = max(plot_df$PredictedAge, na.rm = TRUE),
            label = annotation_text,
            hjust = 0,
            vjust = 1, size = 5
        ) +
        labs(
            title = NULL,
            x = "Gestational Age(weeks)",
            y = paste0("Predicted Age(", c("Knight GA", "Mayne GA")[i], ", weeks)")
        ) +
        theme_bw() +
        theme(
            plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
            plot.subtitle = element_text(hjust = 0.5, size = 12),
            axis.title = element_text(size = 14),
            axis.text = element_text(size = 12)
        )
}

ggpubr::ggarrange(plotlist = gp, nrow = 1, ncol = 2)

## `geom_smooth()` using formula = 'y ~ x'
## `geom_smooth()` using formula = 'y ~ x'

11 Tutorial Example 4: Apply sc-ImmuAging to PBMC scRNA-seq dataset

In this example, we showcase the application of the cell-type-specific clock, scImmuAging, to a PBMC scRNA-seq dataset. The dataset includes 20 samples, profiled to contain only CD4+ and CD8+ T cells.

library(OmniAgeR)
library(Seurat)
library(glmnet)
library(ggplot2)
library(patchwork)
library(ggpubr)
#library(Seurat)
seuratObj <- loadOmniAgeRdata(
    "omniager_yazar_cd4t_cd8t_example",
    verbose = FALSE
)
scImmuAgingOut <- scImmuAging(seuratObj, c("CD4T", "CD8T"))

sc_gp <- list()

for (i in seq_along(scImmuAgingOut)) {
    plot_df <- data.frame(ActualAge = scImmuAgingOut[[i]]$donor$age, PredictedAge = scImmuAgingOut[[i]]$donor$predicted)
    cor_test_result <- cor.test(plot_df$ActualAge, plot_df$PredictedAge)
    correlation <- cor_test_result$estimate
    p_value <- cor_test_result$p.value
    mae <- mean(abs(plot_df$PredictedAge - plot_df$ActualAge))
    p_value_formatted <- ifelse(p_value < 0.001,
        formatC(p_value, format = "e", digits = 2),
        round(p_value, 3)
    )
    annotation_text <- paste0(
        "R = ", round(correlation, 3), "\n",
        "P = ", p_value_formatted, "\n",
        "MAE = ", round(mae, 2)
    )
    sc_gp[[i]] <- ggplot(data = plot_df, aes(x = ActualAge, y = PredictedAge)) +
        geom_point(color = "steelblue", size = 3, alpha = 0.8) +
        geom_smooth(method = "lm", color = "black", se = FALSE) +
        annotate("text",
            x = min(plot_df$ActualAge, na.rm = TRUE),
            y = max(plot_df$PredictedAge, na.rm = TRUE),
            label = annotation_text,
            hjust = 0,
            vjust = 1,
            size = 5
        ) +
        labs(
            title = c("CD4T", "CD8T")[i],
            x = "Chronological Age",
            y = "Predicted Age(sc-ImmuAging)"
        ) +
        theme_bw() +
        theme(
            plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
            plot.subtitle = element_text(hjust = 0.5, size = 12),
            axis.title = element_text(size = 14),
            axis.text = element_text(size = 12)
        )
}

ggpubr::ggarrange(plotlist = sc_gp, nrow = 1, ncol = 2)

12 Tutorial Example 5: Apply Brain_CT_clock to Brain snRNA-seq dataset

In this example, we showcase the application of the brain cell-type-specific aging Clocks to a brain snRNA-seq dataset. The dataset includes 15 donors, profiled to contain only Oligodendrocytes.

library(OmniAgeR)
library(Seurat)
library(glmnet)
library(ggplot2)
library(patchwork)
library(ggpubr)
library(dplyr)

brainSeurat <- loadOmniAgeRdata(
    "omniager_brain_frohlich_control_example_15donors",
    verbose = FALSE
)

# Define cell types of interest
cellTypes <- c("Oligodendrocytes")

# Run all three models for the specified cell types
clockResults <- brainCtClock(
    seuratObj = brainSeurat,
    cellTypes = cellTypes
)

# Use lapply to iterate over each data frame in the list
averagePredictionsDonor <- lapply(clockResults, function(df) {
    df %>%
        group_by(donorId, age, sampleType, celltype) %>%
        summarise(mean_prediction = mean(prediction, na.rm = TRUE))
})

brainPlot <- list()

for (i in seq_along(averagePredictionsDonor)) {
    plot_df <- data.frame(
        ActualAge = averagePredictionsDonor[[i]]$age,
        PredictedAge = averagePredictionsDonor[[i]]$mean_prediction
    )
    cor_test_result <- cor.test(plot_df$ActualAge, plot_df$PredictedAge)
    correlation <- cor_test_result$estimate
    p_value <- cor_test_result$p.value
    mae <- mean(abs(plot_df$PredictedAge - plot_df$ActualAge))
    p_value_formatted <- ifelse(p_value < 0.001,
        formatC(p_value, format = "e", digits = 2),
        round(p_value, 3)
    )
    annotation_text <- paste0(
        "R = ", round(correlation, 3), "\n",
        "P = ", p_value_formatted, "\n",
        "MAE = ", round(mae, 2)
    )
    brainPlot[[i]] <- ggplot(data = plot_df, aes(x = ActualAge, y = PredictedAge)) +
        geom_point(color = "steelblue", size = 3, alpha = 0.8) +
        geom_smooth(method = "lm", color = "black", se = FALSE) +
        annotate("text",
            x = min(plot_df$ActualAge, na.rm = TRUE),
            y = max(plot_df$PredictedAge, na.rm = TRUE),
            label = annotation_text,
            hjust = 0,
            vjust = 1,
            size = 4
        ) +
        labs(
            title = "Oligodendrocytes",
            x = "Chronological Age",
            y = paste0("Predicted Age(", c("SC", "Pseudobulk", "Bootstrap")[i], ")")
        ) +
        theme_bw() +
        theme(
            plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
            plot.subtitle = element_text(hjust = 0.5, size = 12),
            axis.title = element_text(size = 14),
            axis.text = element_text(size = 12)
        )
}

ggpubr::ggarrange(plotlist = brainPlot, nrow = 1, ncol = 3)

13 Tutorial Example 6: Apply Pasta to Brain (MTG) scRNA-seq dataset

In this example, we showcase the application of the Pasta clock to a single-nuclei RNA-seq dataset from the middle temporal gyrus (MTG) (Gabitto et al., 2024). The dataset includes 46 healthy donors and is filtered to contain extratelencephalic projecting glutamatergic cortical neurons.

# Load required libraries
library(OmniAgeR)
library(Seurat)
library(glmnet)
library(magrittr) # For %>% pipe
library(ggplot2)
library(patchwork)
library(ggpubr) # For simplified plot annotations

seu <- loadOmniAgeRdata(
    "omniager_seu_gabitto_2024_filtered",
    verbose = FALSE
)
# Extract and clean chronological age from metadata
seu$age <- seu$development_stage %>%
    gsub("-year.*", "", .) %>%
    gsub("-", " ", .) %>%
    gsub("80 year old and over stage", "85", .)

# Create pseudobulk samples, Pseudobulk construction may follow Salignon et al.
# or be performed using alternative approaches as appropriate
set.seed(42)
seuBulk <- makePseudobulksPasta(
    seu,
    poolBy = c("cell_type", "age"),
    chunkSize = 512,
    verbose = FALSE
)

# Extract the log-normalized expression matrix for prediction
lognormMatrix <- GetAssayData(seuBulk, assay = "RNA", layer = "data")
lognormMatrix <- as.matrix(lognormMatrix)

# Extract the corresponding metadata for the new pseudobulk samples
seuBulkMeta <- seuBulk[[c("chunkSize", "cell_type", "age")]]
seuBulkMeta$age <- as.numeric(seuBulkMeta$age)

# Apply the PASTA clock
# filter_genes = TRUE: Selects only the genes required by the PASTA model.
# rank_norm = TRUE: Applies the rank-normalization required by PASTA.
pastaRes <- pastaScores(lognormMatrix, filterGenes = TRUE, rankNorm = TRUE)

# Prepare the data frame for plotting
plot_df <- data.frame(
    ActualAge = as.numeric(seuBulk$age),
    Prediction = as.numeric(pastaRes$PASTA)
)

# Create the scatter plot
ggplot(data = plot_df, aes(x = ActualAge, y = Prediction)) +
    geom_point(color = "steelblue", size = 3, alpha = 0.8) +
    geom_smooth(method = "lm", color = "black", se = FALSE) +
    ggpubr::stat_cor(
        method = "pearson",
        label.x.npc = "left",
        label.y.npc = "top",
        hjust = 0,
        size = 5
    ) +
    labs(
        title = NULL,
        x = "Chronological Age",
        y = "PASTA Score"
    ) +
    theme_bw() +
    theme(
        plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
        axis.title = element_text(size = 14),
        axis.text = element_text(size = 12),
        legend.position = "none"
    )

14 Tutorial Example 7: Calculate CPR and CHIP Scores

Here, we use DNAm data to calculate the CRP score and CHIP score.

library(OmniAgeR)
hannumBmiqM <- loadOmniAgeRdata(
    "omniager_hannum_example",
    verbose = FALSE
)[[1]]

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

crpRes <- compCRP(hannumBmiqM)

## [OmniAgeR] Retrieving resource: omniager_crp_cpg

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## [OmniAgeR] Successfully loaded 'omniager_crp_cpg'.

## [CRP] Found: 1764, Required: 1765 (Coverage: 99.9%)

## [intCRP] Found: 62, Required: 62 (Coverage: 100.0%)

print(lapply(crpRes, head, 5))

## $CRP
##  GSM990532  GSM990292  GSM989979  GSM989900  GSM990054 
## -0.4065040  0.3002201 -0.1465114 -0.1490234 -0.3857877 
## 
## $intCRP
##   GSM990532   GSM990292   GSM989979   GSM989900   GSM990054 
## -0.29625111  0.23775122 -0.06492197  0.03861159 -0.46205741

chipRes <- compCHIP(hannumBmiqM)

## [OmniAgeR] Retrieving resource: omniager_chip_cpg

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

## [OmniAgeR] Successfully loaded 'omniager_chip_cpg'.

## [AnyCHIP] Found: 9566, Required: 9615 (Coverage: 99.5%)

## [DNMT3A] Found: 5961, Required: 5990 (Coverage: 99.5%)

## [TET2] Found: 5611, Required: 5633 (Coverage: 99.6%)

## [ASXL1] Found: 6025, Required: 6078 (Coverage: 99.1%)

print(lapply(chipRes, head, 5))

## $AnyCHIP
##    GSM990532    GSM990292    GSM989979    GSM989900    GSM990054 
## -0.503068906 -0.359947191  0.006808985 -0.648675130  0.353746991 
## 
## $DNMT3A
##  GSM990532  GSM990292  GSM989979  GSM989900  GSM990054 
## -1.5686495 -0.8673509 -0.1098401 -0.3111652  1.1811087 
## 
## $TET2
##  GSM990532  GSM990292  GSM989979  GSM989900  GSM990054 
## -0.1877774 -0.5414346  0.2673744 -0.5044092  0.1782826 
## 
## $ASXL1
##  GSM990532  GSM990292  GSM989979  GSM989900  GSM990054 
## -0.1334423 -0.5384459  0.0903360 -0.5016612  0.1762066

15 Tutorial Example 8: Get freatures from Various Aging Biomarkers

library(OmniAgeR)
# 1. Get probes for a specific chronological aging predictor
chronologicalProbes <- getMarkerWeights(clockNames = "Horvath2013")

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## loading from cache

print(head(chronologicalProbes$Horvath2013))

##         probe        coef
## 1 (Intercept) 0.695507258
## 2  cg00075967 0.129336610
## 3  cg00374717 0.005017857
## 4  cg00864867 1.599764050
## 5  cg00945507 0.056852418
## 6  cg01027739 0.102862854

16 Tutorial Example 9: Celltype fraction clock based on DNAm

We apply the DNAm Cell-type Fraction (CTF) Clock to 50 blood samples (EPIC array) from the TZH cohort to predict chronological age.

library(OmniAgeR)
library(randomForest)

## randomForest 4.7-1.2

## Type rfNews() to see new features/changes/bug fixes.

## 
## Attaching package: 'randomForest'

## The following object is masked from 'package:ggplot2':
## 
##     margin

library(ggplot2)
library(patchwork)
library(ggpubr)

tzhExample <- loadOmniAgeRdata(
    "omniager_tzh_example_ctf",
    verbose = FALSE
)

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## downloading 1 resources

## retrieving 1 resource

## loading from cache

phenoTypesTzh <- tzhExample[[1]]
tzhFracM <- tzhExample[[2]]
dnamCTFClockOut <- dnamCTFClock(ctfM = tzhFracM)

## [OmniAgeR] Retrieving resource: omniager_dnam_ctf_model

## see ?OmniAgeRData and browseVignettes('OmniAgeRData') for documentation

## downloading 1 resources

## retrieving 1 resource

## loading from cache

## [OmniAgeR] Successfully loaded 'omniager_dnam_ctf_model'.

plot_df <- data.frame(ActualAge = phenoTypesTzh$Age, PredictedAge = dnamCTFClockOut)
cor_test_result <- cor.test(plot_df$ActualAge, plot_df$PredictedAge)
correlation <- cor_test_result$estimate
p_value <- cor_test_result$p.value
mae <- mean(abs(plot_df$PredictedAge - plot_df$ActualAge))
p_value_formatted <- ifelse(p_value < 0.001,
    formatC(p_value, format = "e", digits = 2),
    round(p_value, 3)
)
annotation_text <- paste0(
    "R = ", round(correlation, 3), "\n",
    "P = ", p_value_formatted, "\n",
    "MAE = ", round(mae, 2)
)
ggplot(data = plot_df, aes(x = ActualAge, y = PredictedAge)) +
    geom_point(color = "steelblue", size = 3, alpha = 0.8) +
    geom_smooth(method = "lm", color = "black", se = FALSE) +
    annotate("text",
        x = min(plot_df$ActualAge, na.rm = TRUE),
        y = max(plot_df$PredictedAge, na.rm = TRUE),
        label = annotation_text,
        hjust = 0,
        vjust = 1, size = 5
    ) +
    labs(
        title = NULL,
        x = "Chronological Age",
        y = paste0("Predicted Age(DNAm_CTF)")
    ) +
    theme_bw() +
    theme(
        plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
        plot.subtitle = element_text(hjust = 0.5, size = 12),
        axis.title = element_text(size = 14),
        axis.text = element_text(size = 12)
    )

## `geom_smooth()` using formula = 'y ~ x'

# ggpubr::ggarrange(plotlist = gp, nrow = 1, ncol = 1)

## R version 4.6.0 RC (2026-04-17 r89917)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 24.04.4 LTS
## 
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.23-bioc/R/lib/libRblas.so 
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.12.0  LAPACK version 3.12.0
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_GB              LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## time zone: America/New_York
## tzcode source: system (glibc)
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
## [1] randomForest_4.7-1.2 OmniAgeRData_0.99.3  ggpubr_0.6.3        
## [4] patchwork_1.3.2      ggplot2_4.0.3        OmniAgeR_0.99.4     
## [7] BiocStyle_2.40.0    
## 
## loaded via a namespace (and not attached):
##   [1] RcppAnnoy_0.0.23       splines_4.6.0          later_1.4.8           
##   [4] filelock_1.0.3         tibble_3.3.1           polyclip_1.10-7       
##   [7] preprocessCore_1.74.0  fastDummies_1.7.6      lifecycle_1.0.5       
##  [10] httr2_1.2.2            rstatix_0.7.3          globals_0.19.1        
##  [13] lattice_0.22-9         MASS_7.3-65            backports_1.5.1       
##  [16] magrittr_2.0.5         plotly_4.12.0          sass_0.4.10           
##  [19] rmarkdown_2.31         jquerylib_0.1.4        yaml_2.3.12           
##  [22] httpuv_1.6.17          otel_0.2.0             Seurat_5.5.0          
##  [25] sctransform_0.4.3      spam_2.11-3            sp_2.2-1              
##  [28] spatstat.sparse_3.1-0  reticulate_1.46.0      cowplot_1.2.0         
##  [31] pbapply_1.7-4          DBI_1.3.0              RColorBrewer_1.1-3    
##  [34] abind_1.4-8            Rtsne_0.17             purrr_1.2.2           
##  [37] BiocGenerics_0.58.1    rappdirs_0.3.4         IRanges_2.46.0        
##  [40] S4Vectors_0.50.1       ggrepel_0.9.8          irlba_2.3.7           
##  [43] listenv_0.10.1         spatstat.utils_3.2-3   goftest_1.2-3         
##  [46] RSpectra_0.16-2        spatstat.random_3.4-5  fitdistrplus_1.2-6    
##  [49] parallelly_1.47.0      codetools_0.2-20       tidyselect_1.2.1      
##  [52] shape_1.4.6.1          farver_2.1.2           matrixStats_1.5.0     
##  [55] stats4_4.6.0           BiocFileCache_3.2.0    spatstat.explore_3.8-0
##  [58] Seqinfo_1.2.0          jsonlite_2.0.0         progressr_0.19.0      
##  [61] Formula_1.2-5          ggridges_0.5.7         survival_3.8-6        
##  [64] iterators_1.0.14       foreach_1.5.2          tools_4.6.0           
##  [67] ica_1.0-3              Rcpp_1.1.1-1.1         glue_1.8.1            
##  [70] gridExtra_2.3          mgcv_1.9-4             qs2_0.2.1             
##  [73] xfun_0.57              dplyr_1.2.1            withr_3.0.2           
##  [76] BiocManager_1.30.27    fastmap_1.2.0          digest_0.6.39         
##  [79] R6_2.6.1               mime_0.13              scattermore_1.2       
##  [82] tensor_1.5.1           dichromat_2.0-0.1      spatstat.data_3.1-9   
##  [85] RSQLite_3.52.0         tidyr_1.3.2            generics_0.1.4        
##  [88] data.table_1.18.4      httr_1.4.8             htmlwidgets_1.6.4     
##  [91] uwot_0.2.4             pkgconfig_2.0.3        gtable_0.3.6          
##  [94] blob_1.3.0             lmtest_0.9-40          S7_0.2.2              
##  [97] XVector_0.52.0         htmltools_0.5.9        carData_3.0-6         
## [100] dotCall64_1.2          bookdown_0.46          SeuratObject_5.4.0    
## [103] scales_1.4.0           Biobase_2.72.0         png_0.1-9             
## [106] spatstat.univar_3.1-7  knitr_1.51             reshape2_1.4.5        
## [109] nlme_3.1-169           curl_7.1.0             cachem_1.1.0          
## [112] zoo_1.8-15             stringr_1.6.0          BiocVersion_3.23.1    
## [115] KernSmooth_2.23-26     parallel_4.6.0         miniUI_0.1.2          
## [118] AnnotationDbi_1.74.0   pillar_1.11.1          grid_4.6.0            
## [121] vctrs_0.7.3            RANN_2.6.2             promises_1.5.0        
## [124] stringfish_0.19.0      car_3.1-5              dbplyr_2.5.2          
## [127] xtable_1.8-8           cluster_2.1.8.2        evaluate_1.0.5        
## [130] magick_2.9.1           tinytex_0.59           cli_3.6.6             
## [133] compiler_4.6.0         crayon_1.5.3           rlang_1.2.0           
## [136] future.apply_1.20.2    ggsignif_0.6.4         labeling_0.4.3        
## [139] plyr_1.8.9             stringi_1.8.7          viridisLite_0.4.3     
## [142] deldir_2.0-4           Biostrings_2.80.0      lazyeval_0.2.3        
## [145] spatstat.geom_3.7-3    glmnet_5.0             Matrix_1.7-5          
## [148] ExperimentHub_3.2.0    RcppHNSW_0.6.0         bit64_4.8.0           
## [151] future_1.70.0          KEGGREST_1.52.0        shiny_1.13.0          
## [154] AnnotationHub_4.2.0    ROCR_1.0-12            igraph_2.3.1          
## [157] broom_1.0.13           memoise_2.0.1          RcppParallel_5.1.11-2 
## [160] bslib_0.11.0           bit_4.6.0