Epigenetic age estimation in saliva and in buccal cells

Age estimation based on epigenetic markers is a DNA intelligence tool with the potential to provide relevant information for criminal investigations, as well as to improve the inference of age-dependent physical characteristics such as male pattern baldness or hair color. Age prediction models have been developed based on different tissues, including saliva and buccal cells, which show different methylation patterns as they are composed of different cell populations. On many occasions in a criminal investigation, the origin of a sample or the proportion of tissues is not known with certainty, for example the provenance of cigarette butts, so use of combined models can provide lower prediction errors.In the present study, two tissue-specific and seven age-correlated CpG sites were selected from publicly available data from the Illumina HumanMethylation 450 BeadChip and bibliographic searches, to help build a tissue-dependent, and an age-prediction model, respectively. For the development of both models, a total of 184 samples (N = 91 saliva and N = 93 buccal cells) ranging from 21 to 86 years old were used. Validation of the models was performed using either k-fold cross-validation and an additional set of 184 samples (N = 93 saliva and N = 91 buccal cells, 21–86 years old).The tissue prediction model was developed using two CpG sites (HUNK and RUNX1) based on logistic regression that produced a correct classification rate for saliva and buccal swab samples of 88.59 % for the training set, and 83.69 % for the testing set. Despite these high success rates, a combined age prediction model was developed covering both saliva and buccal cells, using seven CpG sites (cg10501210, LHFPL4, ELOVL2, PDE4C, HOXC4, OTUD7A and EDARADD) based on multivariate quantile regression giving a median absolute error (MAE): ± 3.54 years and a correct classification rate ( %CP±PI) of 76.08 % for the training set, and an MAE of ± 3.66 years and a %CP±PI of 71.19 % for the testing set. The addition of tissue-of origin as a co-variate to the model was assessed, but no improvement was detected in age predictions. Finally, considering the limitations usually faced by forensic DNA analyses, the robustness of the model and the minimum recommended amount of input DNA for bisulfite conversion were evaluated, considering up to 10 ng of genomic DNA for reproducible results. The final multivariate quantile regression age predictor based on the models we developed has been placed in the open-access Snipper forensic classification website.     

Platform-independent models for age prediction using DNA methylation data

Age prediction has been in the spotlight recently because it can provide an important information about the contributors of biological evidence left at crime scenes. Specifically, many researchers have actively suggested age-prediction models using DNA methylation at several CpG sites and tested the candidates using platforms such as the HumanMethylation 450 array and pyrosequencing. With DNA methylation data obtained from each platform, age prediction models were constructed using diverse statistical methods typically with multivariate linear regression. However, because each developed model is based on single-platform data, the prediction accuracy is reduced when applying DNA methylation data obtained from other platforms. In this study, bisulfite sequencing data for 95 saliva samples were generated using massively parallel sequencing (MPS) and compared with methylation SNaPshot data from the same 95 individuals. The predicted age obtained by applying MPS data to an age-prediction model built for methylation SNaPshot data differed greatly from the chronological age due to platform differences. Therefore, novel variables were introduced to indicate the platform type, and construct platform-independent age predictive models using a neural network and multivariate linear regression. The final neural network model had a mean absolute deviation (MAD) of 3.19 years between the predicted and chronological age, and the mean absolute percentage error (MAPE) was 8.89% in the test set. Similarly, the linear regression model showed 3.69 years of MAD and 10.44% of MAPE in the same test set. The platform-independent age-prediction model was made extensible to an increasing number of platforms by introducing platform variables, and the idea of platform variables can be applied to age prediction models for other body fluids.     

DNA methylation-based age prediction from saliva: High age predictability by combination of 7 CpG markers

DNA methylation is currently one of the most promising age-predictive biomarkers. Many studies have reported DNA methylation-based age predictive models, but most of these are based on DNA methylation patterns from blood. Only a few studies have examined age-predictive DNA patterns in saliva, which is one of the most frequently-encountered body fluids at crime scenes. In this study, we generated genome-wide DNA methylation profiles of saliva from 54 individuals and identified CpG markers that showed a high correlation between methylation and age. Because the age-associated marker candidates from saliva differed from those of blood, we investigated DNA methylation patterns of 6 age-associated CpG marker candidates (cg00481951, cg19671120, cg14361627, cg08928145, cg12757011, and cg07547549 of the SST, CNGA3, KLF14, TSSK6, TBR1, and SLC12A5 genes, respectively) in addition to a cell type-specific CpG marker (cg18384097 of the PTPN7 gene) in an independent set of saliva samples obtained from 226 individuals aged 18 to 65 years. Multiplex methylation SNaPshot reactions were used to generate the data. We then generated a linear regression model with age information and the methylation profile from the 113 training samples. The model exhibited a 94.5% correlation between predicted and chronological age with a mean absolute deviation (MAD) from chronological age of 3.13 years. In subsequent validation using 113 test samples, we also observed a high correlation between predicted and chronological age (Spearman’s rho = 0.952, MAD from chronological age = 3.15 years). The model composed of 7 selected CpG sites enabled age prediction in saliva with high accuracy, which will be useful in saliva analysis for investigative leads.