Table 3.

Detailed description of different scaling/normalization approaches used in the benchmarking.

Scaling/normalization method	Single-cell specific	Output containing negative values	Output bounded in [0,1] interval	Reference
Column-wise (= Total count or library size normalization)	No	No	Yes	59
Column min-max	No	No	Yes	60
Column z-score	No	Yes	No	61
Row-wise	No	No	Yes	62
Global min-max	No	No	Yes	60
Global z-score	No	Yes	No	61
Quantile normalization (QN)	No	No	No	63
Upper quartile (UQ)	No	No	No	64
Transcripts per million (TPM)	No	No	No	65
Trimmed mean of M-values (TMM)	No	No	No	66
LogNormalize	No	No	No	67
Median of ratios	No	No	No	13
Scran	Yes	No	No	16
Scater	Yes	No	No	68
Linnorm	Yes	No	No	69
SCTransform (RNBR)	Yes	No	No	15