Statistical strategies and stochastic predictive models for the MARK-AGE data

•The MARK-AGE project aims to develop a prediction model for the biological age.•A proper analysis pipeline is discussed in the light of the state of art.•It is fundamental to use robust estimators that acknowledge the structure of the data.•A train-test split division is necessary to avoid biases in the prediction.•Bayesian methods that allow to include prior medical knowledge should be preferred. MARK-AGE aims at the identification of biomarkers of human aging capable of discriminating between the chronological age and the effective functional status of the organism. To achieve this, given the structure of the collected data, a proper statistical analysis has to be performed, as the structure of the data are non trivial and the number of features under study is near to the number of subjects used, requiring special care to avoid overfitting. Here we described some of the possible strategies suitable for this analysis. We also include a description of the main techniques used, to explain and justify the selected strategies. Among other possibilities, we suggest to model and analyze the data with a three step strategy: