Table 2.

Input data type and parameter setting of different imputation methods

Algorithm	Version	Input data type${}^{\ast }$	Parameter Setting
SIMLR [15]	0.1.3	Raw count, TPM	Default
ZINBWaVE [6]	1.6.0	Raw count	Default
scImpute [7]	0.0.9	Raw count	‘Kcluster’ was set to 5 for simulated datasets, 20 for GSE123813 and 10 for the others.${}^{\ast \ast }$
DrImpute [16]	1.0	Raw count, TPM	‘ks’ was set to 5:10
SAVER [8]	1.1.1	Raw count, TPM	Default
MAGIC [13]	1.5.2	Raw count, TPM	Default
NE [14]	–	Raw count, TPM	Default
scVI [11]	0.3.0	Raw count, TPM	‘new_n_genes’ was set to the number of genes of each dataset.
DCA [12]	0.2.2	Raw count	Default
scScope [9]	0.1.5	Raw count, TPM	Default
SAUCIE [10]	–	Raw count, TPM	Default

${}^{\ast }$For scImpute, ZINBWaVE and DCA, only raw counts are allowed for input

${}^{\ast \ast }$To ensure that scImpute obtained the same prior knowledge as other methods, we didn’t provide the accurate number of cell types for it