The aging process is accompanied by changes in DNA methylation that has become one of the important biomarkers of aging. Research in the field of aging has become more and more hot in recent years. Age prediction helps study biological aging the research of biological aging, but the prediction accuracy needs to be further improved. Most of the previous studies were based on linear regression models that used highly correlated CpG loci in DNA methylation data as features to predict age. Compared with machine learning models, deep learning models is more inclusive for multi-feature tasks and can select more CpG loci as features. In the methylation data of Illumina 27K and Illumina 450K arrays, methylation data of 21 368 CpG sites were selected as input. MLPAge, a pan-tissue age prediction method, was established using multi-layer perceptron to predict age. MLPAge was compared with the Horvath 353 CpG clock, the standard in the pan-tissue age prediction methods, industry in an independent validation set of 2 310 samples from 8 studies with Median Absolute Deviation (MAD) of 3.77 years. It was found that the multi-layer perceptron is able to better extract age-related features and has higher accuracy in age prediction, providing a new deep learning-based solution for the filed.