Skip to main content

Table 2 Example simulations illustrate the limitations of current techniques

From: Analysis and correction of compositional bias in sparse sequencing count data

Net compositional change (Λg) Average sample depth CLR TMM CSS Scran W 0 W 1 W 2 W 3
36.86X 1M 1.36 1.45 5.41 22.57 19.32 31.44 30.65 32.01
7.75X 10K .95 3.05 1.47 12.08 (14/40 samples failed) 5.30 6.32 6.31 6.70
  1. Shown are the group-wise true and reconstructed compositional scales from the methods compared on two simulated examples, each at different sequencing depths and at different total true absolute abundance changes for a roughly 54K features with control group proportions derived from the Lung microbiome. Low-coverage and/or high compositional changes are problematic for current techniques due to the sparsity they cause in the count data. W1,…W3 are Wrench estimators proposed in the Methods section that adjust the base estimator W0 for feature-wise zero-generation properties. All are presented here for comparison purposes. Our default estimator is W2