Skip to main content

Table 1 Synthetic example to illustrate the effect of addition of pseudo-counts over the estimation of LFCs in overall expression

From: Beyond comparisons of means: understanding changes in gene expression at the single-cell level

 

Empirical

Adding 0.5

Adding 1

 

estimate

pseudo-counts

pseudo-count

Overall expression rate in population 1 (\(\mu ^{(1)}_{i}\))

10

10.5

11

Overall expression rate in population 2 (\(\mu ^{(2)}_{i}\))

0

0.5

1

LFC in overall expression 1 vs 2

+

3.04

2.40

  1. For simplicity, we assume that normalization is not required so that pseudo-counts are linearly reflected in the overall expression rates. While pseudo-counts introduce an additive effect, LFC estimates measure changes on a multiplicative scale. Hence, addition of pseudo-counts leads to an artificial deflation of LFC estimates. As a consequence, such estimates cannot be meaningfully interpreted