Association of DNA methylation age acceleration with digital clock drawing test performance: the Framingham Heart Study

Background: Cognitive function measured by digital clock drawing test (dCDT) has drawn attention for their precision, automation, and reproductivity. However, the relationship between digital cognitive metrics and biological aging is lacking.

We conducted association analyses between cognitive function measured by dCDT and biological aging metrics quantified by five DNA methylation (DNAm) age metrics (Horvath, Hannum, GrimAge, PhenoAge, and DunedinPACE) in the Framingham Heart Study (FHS). We conducted linear regression to investigate the association between cognitive functions (global cognitive function and four sub-domain functions) and DNAm age acceleration, adjusting for covariates. We used a false discovery rate (FDR) < 0.05 for significance.

Among the 1,798 FHS participants (mean age 65±13, 53% women), we found that a lower dCDT total score is associated with DNAm age acceleration. Larger magnitudes of associations were observed in older participants (≥ 65 years). The dCDT total score showed the strongest association with the DundinPACE in the pooled sample (beta = -2.1, FDR = 0.0004), the younger (beta = -1.9, FDR = 0.02), and older age group (beta = -2.2, FDR = 0.01). The dCDT total score was significantly associated with age acceleration estimated by Horvath (beta=-1.9, FDR =0.01) and PhenoAge (beta=-2.5, FDR=0.01) in older participants while not in the pooled sample or younger participants (<65 years). In sub-domain cognitive functions, we found that simple motor function was significantly associated with DunedinPACE (FDR = 0.005) in both age groups and associated with GrimAge (FDR = 0.05 in older age group), indicating the deterioration in various organ systems may particularly impact this domain. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint

Our findings suggest that cognitive function measured by a digital clock drawing test is associated with DNAm age acceleration in middle-aged and older participants in the FHS, potentially shedding light on the epigenetic mechanisms underlying digitally measured cognitive function. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint

Cognitive function, representing a crucial aspect of overall brain health, encompasses various domains such as memory, reasoning, and attention1. Neuropsychological (NP) tests are typically used to measure cognitive functions for individuals, focusing on one or several specific cognitive domains. For example, the traditional clock drawing test evaluates spatial dysfunction and neglect2, conducted with pen on paper. However, the assessment of the NP test is often biased. The use of computerized devices in NP tests is gaining recognition for their efficiency, precision, automation, and reproductivity nature3. The Digital Clock Drawing test (dCDT), an extension of the traditional clock drawing test, evaluates the overall cognitive function and specific sub-domains such as motor function, memory, spatial reasoning, and information processing4. Previous research has demonstrated strong associations of dCDT performance with mild cognitive impairment5, and with clinical indices of neurodegeneration, such as brain volume6 and NP tests5, showing a comparable performance between dCDT and other NP tests. Compared to traditional NP tests, dCDT can capture more subtle cognitive changes7. Aging is an inevitable process for humans, resulting in a decline in physiological capacity and an increasing risk of various disease conditions8, including neurodegenerative diseases9. Biological aging, the changes at the molecular level10,11, is essential to understand the heterogeneity in healthy aging12. Epigenetic modifications, such as DNA methylation (DNAm), measure biological aging at the epigenetic level13-15. DNAm age16 has been associated with general health17,18 and neurodegenerative diseases19. The first generation of epigenetic clocks use methylation levels across varying numbers of 5'-C-phosphate-G-3'(CpG) sites with a few clinical markers to predict chronological age16,20. Subsequent generations of epigenetic clocks have incorporated additional clinical measurements to enhance their accuracy for age prediction21-23. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint DNAm age acceleration is assessed by regressing the estimated DNAm age on chronological age24. Studies have established correlations between traditional cognitive assessments and DNAm age acceleration25, showing the possibility that DNAm age acceleration is a risk factor or marker for cognitive function decline. However, research on the relationship between cognitive function measured by digital devices and DNAm age acceleration is currently lacking. Our study aimed to fill in this gap. We hypothesize that cognitive function measured by digital devices is associated with DNAm aging. We investigated the associations of overall cognitive function and specific sub-domains measured by DCTclock with several DNAm-based age acceleration metrics in the Framingham Heart Study (FHS). We seek to gain additional insights into the underlying molecular mechanism of cognitive function measured by digital devices.

Study populations FHS, initiated in 1948, is an epidemiological prospective cohort to study risk factors for CVD26. All FHS cohorts, including the original cohort (Gen1), offspring cohort (Gen2), and third- generation cohort (Gen3), have undergone routine health examinations every two to six years. Our study included 1,264 Gen2 participants at exam 8 (2005-2008) and 688 Gen3 participants at exam 2 (2008-2011). We excluded participants whose blood samples were not collected at exam or who did not attend dCDT (2011 -2018). After excluding participants with covariates (e.g., age, gender, education, cell counts) missing, we included 1,789 participants (Figure 1) in our . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint statistical analyses. For participants with multiple dCDT tests, we selected dCDT data closest to DNA methylation measurement dates for inclusion in our study. DCTclock DCTclock serves as an FDA-approved automated screening tool for detecting cognitive changes4. Data collection for dCDT using DCTclock involved utilizing computerized neuropsychological assessment devices: digital pen and digital paper. Following the standard protocol, participants were instructed to draw clocks with the command '10 after 11' and then replicated another clock by copying a provided model27,28. Both the drawing process and the final drawing results were recorded, capturing spatial and temporal data, and analyzed through the DCTclock pipeline. These data were treated as input to a trained convolutional neural network to recognize individual symbols with classification probability (e.g., clock face, digits, and small noise stocks). After classifying the individual symbols in drawing, these symbols were used to derive various measurements, such as the correct placement of clock components and pen speed. These measurements were organized into four groups representing different cognitive aspects: Drawing Efficiency, Simple and Complex Motor, Information Processing, and Spatial Reasoning. Drawing efficiency, for instance, evaluated the efficiency in terms of the time spent on drawing and the size of the drawing. Similarly, Simple and Complex Motor measurements represented motor and non-motor cognitive functions, including maximum movement speed. Information Processing focused on cognitive functions like thinking time and latencies, while Spatial Reasoning focused on spatial abilities through geometric property measurements. A composite score was calculated for each cognitive aspect mentioned above using a Lasso regularized logistic regression model4, incorporating all the measurements as parameters. Given . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint the performance of two tasks (command and copy), eight sub-domain scores were generated. Additionally, a dCDT total score was computed using a Lasso logistic regression model4. Both sub-domain scores and the total score range from 0 to 100. DNAm measurements DNAm adds a methyl group onto the 5th carbon of cytosine to form 5-methylcytosine29. DNAm measurements were conducted using whole blood samples collected during exam 8 for Gen2 and exam 2 for Gen3. DNAm profiling was carried out through a series of procedures, including bisulfite conversion, whole genome amplification, fragmentation, array hybridization, and single-base pair extension30. The Illumina Human Methylation 450K Bead chips (Illumina Inc., San Diego, CA) were employed to analyze the DNA methylation levels across three different laboratories. Detailed information regarding DNAm quantification and quality control procedures in FHS had been previously documented31. DNAm aging metrics DNAm age is an estimator of aging based on DNAm patterns. Three generations of epigenetic clocks were calculated. The first generation, Horvath’s age16 and Hannum’s age20, utilize a set of CpG sites to estimate DNAm age. Horvath’s age calculated a weighted average of 353 clock CpGs with a calibration function to estimate aging from multiple tissues. Hannum’s age considered 71 CpG sites along with a few clinical parameters (gender, BMI, etc.) to predict aging in whole blood samples. GrimAge22 and PhenoAge23 are the second-generation DNAm aging metrics. Both methods calculated the DNAm age by integrating methylation levels with clinical . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint markers. GrimAge first utilized DNAm data to estimate each plasma protein biomarker and smoking pack years. Then, 7 DNAm-based plasma protein biomarkers and DNAm-based smoking pack year were selected, which included 1,030 unique CpG sites along with gender and chronological age to predict time to death22. PhenoAge selected 513 CpGs to predict a linear combination of chronological age and nine clinical markers (e.g., Albumin, White blood cell count), which predicted the time to death23. We employed the principal component version of epigenetic clocks (PC-based clocks) to minimize unobserved technical confounders32. The DNAm age acceleration for the first- and second-generation aging metrics were residuals calculated by regressing each DNAm age on chronological age. Residuals larger than zero will be considered as accelerated aging. The third generation, DunedinPACE, differed from the previous generations by predicting the pace of aging per year rather than age in years21. The pace of aging was calculated from 173 CpGs based on longitudinal change of 19 clinical biomarkers (e.g., blood pressure, total cholesterol, blood urea nitrogen), representing an average rate of biological aging per year of 1-year chronological age21. This pace of aging was used as DNAm age acceleration for the following analysis. Covariates Covariates included the age at the dCDT, the time interval between the dCDT and blood sample collection, gender, educational level, cell count information, and family relationship. The time interval was computed as the age at the dCDT minus the age at blood sample collection. Educational levels were categorized into four groups: less than high school completion, high school graduate, some college, and college graduate. Cell count information was derived from DNAm data. We included the count number of Cytotoxic T cells (CD8+T), B lymphocytes . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint (CD19+ B), granulocytes (Gran), monocytes (Mono), Natural killer cells (NK), and Helper T Cells (CD4+T) as covariates. Family relationships were included as random effects in the model. Statistical analysis The primary analysis explored the relationship between the dCDT total score as the outcome variable and various DNAm aging metrics as the predictor variables. The residuals were computed by regressing the DNAm age metrics on chronological age to obtain DNAm age acceleration. The residuals were used as DNAm age acceleration in the following analysis. To facilitate interpretation, we standardized the DNAm age residuals to have a mean of 0 with a standard deviation (SD) of 1. Participants were stratified into two age groups (<65 and ≥65 years), using age at blood sample collection, to address for age modification effects in the associations. Linear mixed models were employed to assess the association between the dCDT total score and DNAm aging metrics. We adjusted for age at the dCDT, gender, and educational level and used family as a random effect in both combined samples and age-stratified analyses. To investigate the association between dCDT sub-domain scores and DNAm aging metrics, linear mixed models were employed with the same set of covariates. The False Discovery Rate (FDR) method33 was applied to adjust for multiple testing34.

Participant characteristics . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint This study included 1,789 middle-aged and older participants in FHS (mean age 65 ±13 at dCDT, 53% women) (Table 1). On average, DNAm was measured seven years before the dCDT measurement (Supplementary Figure 1). Education levels were significantly higher in the younger age group (<65 years) compared to the older age group (≥65 years). On average, participants in the younger age group exhibited higher dCDT total score and sub-domain scores compared to those in the older age group (all p <0.001). The average estimated DNAm ages differed between DNAm metrics (Supplemental Table 2). For instance, the mean Hannum age was estimated at 62, while the Horvath age was estimated at 53 in pooled samples. Compared to the mean chronological age, the mean DNAm ages calculated by Hannum, GrimAge, and DunedinPACE showed acceleration (with DNAm age higher than chronological age) in the pooled sample, whereas PhenoAge and Horvath showed lower mean DNAm ages than the mean chronological age. We further observed that, except for DunedinPACE, males tended to have advanced DNAm ages compared to females (Supplemental Table 3). Association between dCDT total score and DNA methylation age acceleration In the pooled sample, we observed that a higher dCDT total score was associated with lower DNAm age estimated by all DNAm aging metrics (Figure 2). However, the association was significant only for DunedinPACE, where a one-SD higher level in the pace of aging was associated with a 2.1-unit lower level in the dCDT total score (FDR=0.0004) (Figure 2). Age showed a significant effect modification of the association (p = 0.004) with DunedinPACE but not for other epigenetic age metrics (Supplemental Table 4). Thus, we conducted stratified analyses by age groups for all DNAm age metrics. In the younger age group (<65 years), DunedinPACE was the only DNAm metric associated with the dCDT total score (beta = -1.9, . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint FDR = 0.02). In contrast, in the older age group, three DNAm aging metrics were significantly associated with the dCDT total score: DunedinPACE (beta=-2.2, FDR=0.01), DNAm age acceleration estimated by Horvath (beta=-1.9, FDR=0.01) and PhenoAge (beta=-2.5, FDR=0.01). Although the association between dCDT total score and other DNAm aging metrics was not significant, the directionality of the associations was consistent in both older and younger age groups (Figure 3, Supplementary Figure 3). Association between dCDT sub-domain scores and DNA methylation age acceleration In the pooled sample, we observed a significant association between DunedinPACE and both the dCDT simple motor function score in the command task (FDR = 0.01) and the spatial reasoning score from the copy task (FDR = 0.01). A one-SD higher level in the pace of aging was associated with a 0.7-unit decrease in the dCDT simple motor function score and a 1.6-unit decrease in the spatial reasoning score. GrimAge was also found to be significantly associated with the dCDT simple motor function score in both the copy task (beta = -0.7, FDR = 0.005) and command task (beta=-0.9, FDR = 0.005) in the pooled sample. No other significant association was found between the other epigenetic age acceleration metrics and the dCDT sub-domain score in the pooled sample (Figure 2). Different from the result in the pooled sample, we only observed similar results for GrimAge in the older age group (>65 years), where a one-SD increase in GrimAge is significantly associated with a 1.1-unit decrease in dCDT simple motor function score in command task (FDR = 0.04) (Supplementary Figure 2). In addition, PhenoAge was significantly associated with the dCDT simple motor function score in the command task (beta =-1.0, FDR = 0.04). While no significant association was found between dCDT sub-domain . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint scores and DNAm age in the younger age group, the directionality was consistent for most associations in both younger and older age groups (Supplementary Figure 2, 3).

In this study, we investigated the association between cognitive function measured by the dCDT and DNAm aging metrics in 1,789 middle-aged and older participants in the FHS. We found that lower dCDT total scores were consistently associated with advanced biological age quantified by DNAm aging metrics. Among all DNAm aging metrics, DunedinPACE showed significant association with the dCDT total score in the pooled sample, younger (<65 years), and older (≥ 65 years) age groups. In contrast, several other DNAm aging metrics showed significant associations with the dCDT total score only in the older participant group. As the need for early diagnosis in the preclinical stage of AD grows, advanced screening tools for cognition are desired to capture subtle cognitive changes35. The dCDT has drawn attention for its efficiency, precision, automation, and reproductivity3. It has been demonstrated to be effective in distinguishing cognitive impairment from normal function4,7. Additionally, it is comparable to established NP tests, such as the Wechsler Memory Scale and Boston Naming Test, in discriminating between participants with mild cognitive impairment (MCI)6 and those with normal function. Unlike traditional paper and pen tests, the dCDT captures more granular data, including visuospatial, time-base, and kinematic details, offering a more detailed assessment. Its automated scoring system provides clinicians with objective and interpretable results4. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint DNAm is a crucial epigenetic modification that influences gene expression without altering the underlying DNA sequence29. By incorporating different sets of CpG sites, the three generations of epigenetic age metrics offer varied perspectives on biological aging. Unlike the first- and second-generation clocks, which primarily estimate DNAm aging based on chronological age, clinical markers and time to death16,20, the third-generation clock, DunedinPACE, focuses on a set of different clinical markers (representing the progressive decline across multiple organ systems captured by multiple CpG sites)21. In our study, DunedinPACE is negatively associated with overall cognitive function in both younger and older age groups, while other age clocks only showed significant results in the older age group. These findings indicate that global cognition function, which is an indicator of brain health, may reflect the aging rates of multiple organ systems. The analysis of dCDT sub-domain scores with DNAm age provides additional insights into how epigenetic aging may influence specific cognitive domains. For instance, the simple motor function and spatial reasoning sub-domain scores show significant associations with DunedinPACE, indicating that the deterioration in various organ systems may particularly impact these domains. GrimAge is based on seven plasma protein markers related to various diseases and conditions, including cardiovascular disease (Plasma B2M) and cognitive functions (Plasma B2M, ADM, cystatin C, and leptin)16. Our observation that simple motor function was associated with advanced GrimAge age indicates that the decline of simple motor functions might be related to abnormal levels of these protein markers in the plasma. Future studies are necessary to investigate the associations of these plasma proteins with cognitive decline. Digital cognitive measures displayed stronger associations with most DNAm aging metrics among older (≥65 years) compared to younger (<65 years) participants, likely to reflect . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint the cumulative and nonlinear age influences on both brain health and DNAm. This is consistent with our earlier findings of stronger associations between alcohol consumption and epigenetic age metrics38. For instance, overall cognitive function exhibited significant associations with PhenoAge in the older age group while not in the younger age group. In contrast, the global cognitive function score was associated with DunedinPACE in both age groups, indicating that DunedinPACE might be more sensitive to capturing the subtle influence of variations in the pace of aging on cognitive changes. Several similar studies have investigated the association of cognitive function measured by traditional methods with DNAm aging metrics. Marioni et al.39 reported that general cognitive ability was associated with Horvath age in participants over the age of seventy. Our findings, using the dCDT, are consistent with this, showing similar associations with total score in older participants (65 years and above). Another study assessed how various cognitive tests (e.g., MMSE, ADAS-Cog-13, MoCA) associated with DNAm age acceleration in participants with a mean age of seventy-five25. They found that the faster pace of aging measured by DunedinPACE correlates with more severe cognitive decline25. Our results align with this, showing that worse cognitive function was associated with DunedinPACE, with a larger magnitude in older participants. Furthermore, our findings that higher PhenoAge and DunedinPACE were associated with poorer performance scores in older participants were also consistent with the previous findings, where they found that higher PhenoAge and DunedinPACE are associated with worse cognitive performance25. Our study has limitations, including a lack of diversity (all participants were Whites) in our study sample. Further research with diverse groups is needed. In addition, there is an . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint approximate 7-year gap between DNA methylation measurement and dCDT. This may introduce biases due to changes in DNA methylation levels and clinical conditions. Our study has several strengths. We employed the dCDT to measure cognitive function, which is a novel measure for assessing cognitive function. In addition, we employed PC-based clocks, which use principal components to reduce noise and enhance accuracy. To mitigate multiple testing, we applied FDR adjustment, which is more appropriate than Bonferroni correction for the presence of correlated outcome and predictor variables. In conclusion, our study investigated how digital cognitive function assessed by dCDT relates to biological aging, measured by DNAm. These findings highlight the potential role of DNA methylation in cognitive function. Further research is needed to uncover the underlying biological pathways behind this association, particularly in more diverse populations. Fundings This work was supported by the National Heart, Lung, and Blood Institute contract (N01-HC- 25195; HHSN268201500001I) and grants from the National Institute on Aging (AG008122, AG062109, AG068753). Conflict of Interest Dr. Au is a scientific advisor to Signant Health and Novo Nordisk. The other authors have no conflict of interest. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint Table 1. Demographic and Clinical characteristics of study participants Total Age below 65 Age above or equal to 65 P-value (n=1789) (n=804) (n=985) Female, n (%) 955 (53 %) 407 (51 %) 548 (56 %) 0.208 Age at dCDT (years) 65 (± 13) 54 (± 8) 75 (± 7) Age at DNAm (years) 58 (± 12) 48 (± 7) 67 (± 7) Education levels, n (%) Incomplete high school 30 (2 %) 6 (1 %) 24 (2 %) <0.001 High school graduate 332 (19 %) 108 (13 %) 224 (23 %) Some college 556 (31 %) 231 (29 %) 325 (33 %) College graduate or above 871 (49 %) 459 (57 %) 412 (42 %) Dementia 51 (3 %) 0 (0 %) 51 (5 %) <0.001 Note: In this study, we used age at DNAm for age group stratification. Mean with standard deviation (SD) was provided for a continuous variable, while count and proportion were provided for categorical variables. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint Figure 1. Flow chart of study design. The participation of dCDT was solely based on consent. In the Framingham Heart Study, we identified participants with digital clock drawing test (dCDT) measurements and DNA methylation (DNAm). Five DNAm age metrics were calculated. Epigenetic (DNAm) age acceleration (EAA) was calculated by regression of the DNAm age metrics on chronological age. Primary analysis focused on the association between dCDT total scores and EAAs, whereas secondary analysis focused on the association between dCDT sub-domain scores and EAAs. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint Figure 2. Association between dCDT scores and DNAm age acceleration in 1789 participants of the Framingham Heart Study. The dCDT total score includes command task composite scores and copy task composite scores. DNA methylation age acceleration was obtained by regressing DNAm age metrics on chronological age. We conducted association analysis between standardized DNAm age acceleration and the dCDT total score, adjusted for age, gender, and education. The numbers inside each cell represent the P-values of the associations. The color represents the change in dCDT scores corresponding to a one SD increase in DNAm age acceleration. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint Figure 3. Comparison of effect size in the association between DNAm age acceleration and the dCDT total score. DNAm age acceleration was obtained by regressing DNA methylation (DNAm) metrics on chronological aging, followed by standardization with a mean of zero and SD of one. We conducted an association analysis between the dCDT total score and standardized DNAm age acceleration. CI, confidence interval. Unit change in dCDT total score in response to one SD increase of EAA Epigenetic age acceleration Effect size (95%CI) . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint

1. Fisher GG, Chacon M, Chaffee DS. Chapter 2 - Theories of Cognitive Aging and Work. In: Baltes BB, Rudolph CW, Zacher H, eds. Work Across the Lifespan. Academic Press; 2019:17-45. 2. Agrell B, Dehlin O. The clock-drawing test. Age and Ageing. 2012;41(suppl_3):iii41-iii45. doi:10.1093/ageing/afs149 3. Bauer RM, Iverson GL, Cernich AN, Binder LM, Ruff RM, Naugle RI. Computerized Neuropsychological Assessment Devices: Joint Position Paper of the American Academy of Clinical Neuropsychology and the National Academy of Neuropsychology†. Archives of Clinical Neuropsychology. 2012;27(3):362-373. doi:10.1093/arclin/acs027 4. Souillard-Mandar W, Penney D, Schaible B, Pascual-Leone A, Au R, Davis R. DCTclock: Clinically- Interpretable and Automated Artificial Intelligence Analysis of Drawing Behavior for Capturing Cognition. Frontiers in Digital Health. 2021;3doi:10.3389/fdgth.2021.750661 5. Yuan J, Libon DJ, Karjadi C, et al. Association Between the Digital Clock Drawing Test and Neuropsychological Test Performance: Large Community-Based Prospective Cohort (Framingham Heart Study). J Med Internet Res. 2021/6/8 2021;23(6):e27407. doi:10.2196/27407 6. Yuan J, Au R, Karjadi C, et al. Associations Between the Digital Clock Drawing Test and Brain Volume: Large Community-Based Prospective Cohort (Framingham Heart Study). Journal of Medical Internet Research. 2022;24(4):e34513. doi:10.2196/34513 7. Davoudi A, Dion C, Amini S, et al. Classifying Non-Dementia and Alzheimer’s Disease/Vascular Dementia Patients Using Kinematic, Time-Based, and Visuospatial Parameters: The Digital Clock Drawing Test. Journal of Alzheimer's Disease. 2021;82:47-57. doi:10.3233/JAD-201129 8. Troen BR. The biology of aging. Mt Sinai J Med. Jan 2003;70(1):3-22. 9. Hou Y, Dan X, Babbar M, et al. Ageing as a risk factor for neurodegenerative disease. Nature Reviews Neurology. 2019/10/01 2019;15(10):565-581. doi:10.1038/s41582-019-0244-7 10. Hayflick L. Theories of biological aging. Experimental Gerontology. 1985/01/01/ 1985;20(3):145- 159. doi:https://doi.org/10.1016/0531-5565(85)90032-4 11. Baker GT, 3rd, Sprott RL. Biomarkers of aging. Exp Gerontol. 1988;23(4-5):223-39. doi:10.1016/0531-5565(88)90025-3 12. Lowsky DJ, Olshansky SJ, Bhattacharya J, Goldman DP. Heterogeneity in healthy aging. J Gerontol A Biol Sci Med Sci. Jun 2014;69(6):640-9. doi:10.1093/gerona/glt162 13. Han S, Brunet A. Histone methylation makes its mark on longevity. Trends Cell Biol. Jan 2012;22(1):42-9. doi:10.1016/j.tcb.2011.11.001 14. Maegawa S, Hinkal G, Kim HS, et al. Widespread and tissue specific age-related DNA methylation changes in mice. Genome Res. Mar 2010;20(3):332-40. doi:10.1101/gr.096826.109 15. López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G. The Hallmarks of Aging. Cell. 2013;153(6):1194-1217. doi:10.1016/j.cell.2013.05.039 16. Horvath S. DNA methylation age of human tissues and cell types. Genome Biology. 2013;14(10):R115. doi:10.1186/gb-2013-14-10-r115 17. Boks MP, Derks EM, Weisenberger DJ, et al. The Relationship of DNA Methylation with Age, Gender and Genotype in Twins and Healthy Controls. PLOS ONE. 2009;4(8):e6767. doi:10.1371/journal.pone.0006767 18. Ryan J, Wrigglesworth J, Loong J, Fransquet PD, Woods RL. A Systematic Review and Meta- analysis of Environmental, Lifestyle, and Health Factors Associated With DNA Methylation Age. The Journals of Gerontology: Series A. 2020;75(3):481-494. doi:10.1093/gerona/glz099 19. Landgrave-Gómez J, Mercado-Gómez O, Guevara-Guzmán R. Epigenetic mechanisms in neurological and neurodegenerative diseases. Review. Frontiers in Cellular Neuroscience. 2015- February-27 2015;9doi:10.3389/fncel.2015.00058 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint 20. Hannum G, Guinney J, Zhao L, et al. Genome-wide Methylation Profiles Reveal Quantitative Views of Human Aging Rates. Molecular Cell. 2013;49(2):359-367. doi:10.1016/j.molcel.2012.10.016 21. Belsky DW, Caspi A, Corcoran DL, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. Elife. Jan 14 2022;11doi:10.7554/eLife.73420 22. Lu AT, Quach A, Wilson JG, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging. 2019;11(2):303-327. doi:10.18632/aging.101684 23. Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging. 2018;10(4):573-591. doi:10.18632/aging.101414 24. Zavala DV, Dzikowski N, Gopalan S, et al. Epigenetic Age Acceleration and Chronological Age: Associations With Cognitive Performance in Daily Life. The Journals of Gerontology: Series A. 2023;79(1)doi:10.1093/gerona/glad242 25. Sugden K, Caspi A, Elliott ML, et al. Association of Pace of Aging Measured by Blood-Based DNA Methylation With Age-Related Cognitive Impairment and Dementia. Neurology. Sep 27 2022;99(13):e1402-e1413. doi:10.1212/wnl.0000000000200898 26. Dawber TR, Meadors GF, Moore FE, Jr. Epidemiological approaches to heart disease: the Framingham Study. Am J Public Health Nations Health. Mar 1951;41(3):279-81. doi:10.2105/ajph.41.3.279 27. Price CC, Cunningham H, Coronado N, et al. Clock Drawing in the Montreal Cognitive Assessment: Recommendations for Dementia Assessment. Dementia and Geriatric Cognitive Disorders. 2011;31(3):179-187. doi:10.1159/000324639 28. Piers RJ, Devlin KN, Ning B, et al. Age and Graphomotor Decision Making Assessed with the Digital Clock Drawing Test: The Framingham Heart Study. Journal of Alzheimer's Disease. 2017;60:1611- 1620. doi:10.3233/JAD-170444 29. Moore LD, Le T, Fan G. DNA Methylation and Its Basic Function. Neuropsychopharmacology. 2013/01/01 2013;38(1):23-38. doi:10.1038/npp.2012.112 30. Wang M, Li Y, Lai M, et al. Alcohol consumption and epigenetic age acceleration across human adulthood. Aging. 2023;doi:10.18632/aging.205153 31. Liu C, Marioni RE, Hedman ÅK, et al. A DNA methylation biomarker of alcohol consumption. Molecular Psychiatry. 2018;23(2):422-433. doi:10.1038/mp.2016.192 32. Higgins-Chen AT, Thrush KL, Wang Y, et al. A computational solution for bolstering reliability of epigenetic clocks: Implications for clinical trials and longitudinal tracking. Nat Aging. Jul 2022;2(7):644- 661. doi:10.1038/s43587-022-00248-2 33. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodological). 1995;57(1):289-300. doi:https://doi.org/10.1111/j.2517-6161.1995.tb02031.x 34. Jafari M, Ansari-Pour N. Why, When and How to Adjust Your P Values? Cell J. Jan 2019;20(4):604-607. doi:10.22074/cellj.2019.5992 35. Laske C, Sohrabi HR, Frost SM, et al. Innovative diagnostic tools for early detection of Alzheimer's disease. Alzheimers Dement. May 2015;11(5):561-78. doi:10.1016/j.jalz.2014.06.004 36. Salameh Y, Bejaoui Y, El Hajj N. DNA Methylation Biomarkers in Aging and Age-Related Diseases. Review. Frontiers in Genetics. 2020-March-10 2020;11doi:10.3389/fgene.2020.00171 37. Johnson AA, Akman K, Calimport SR, Wuttke D, Stolzing A, de Magalhães JP. The role of DNA methylation in aging, rejuvenation, and age-related disease. Rejuvenation Res. Oct 2012;15(5):483-94. doi:10.1089/rej.2012.1324 38. Wang M, Li Y, Lai M, et al. Alcohol consumption and epigenetic age acceleration across human adulthood. Aging (Albany NY). Oct 26 2023;15(20):10938-10971. doi:10.18632/aging.205153 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 7, 2024. ; https://doi.org/10.1101/2024.11.06.24316862doi: medRxiv preprint