In the paper we have constructed a XGBoost model with capacities to classify cell from moderate and severe COVID-19 patients from their scRNASeq profiles. All the high-throughput data were obtained from (Liao et al., 2020). With the purpose to assess the capacities of this machine learning techniques, we have constructed two models. One of the models have taked into account the entire set of genes that were reported in the original paper. The second model has reduced the set genes by excluding those genes associated with quality factors. We have organized the material into two folders. One folder, XGBoost_Model, includes the model used to reconstructed the results shown in the main paper. The second folder, Reconstruct_XGBoost_Model, give more detail of how models (complete and reduced) were reconstructed. This last folder gives details of how the data were preprocessed and how the model was training and testing.