Identification of gene and protein signatures associated with long-term effects of COVID-19 on the immune system after patient recovery by analyzing single-cell multi-omics data using a machine learning approach