In this way, we created downsampled datasets of 95,316 TCs, 35,254 NKs, 22,042 BCs, 39,144 MCs and 8,244 DCs in cohort-1 and 57,910 TCs, 34,857 NKs, 13,812 BCs, 45,431 MCs and 7,990 DCs in cohort-2 for analysis.