mbImpute: an accurate and robust imputation method for microbiome data

Table 2 Effectiveness of mbImpute in identifying zeros due to downsampling of Qin et al.’s T2D WGS dataset [19]. For each of two removal rates 40% and 70%, we repeat independent downsampling for ten times

Removal rate	40%	70%
% of downsampling zeros identified	95.83%±0.46%	92.83%±0.92%
Pearson correlation before imputation	0.7565±0.0023	0.5261±0.0016
Pearson correlation after imputation	0.8747±0.0100	0.7582±0.0235

For each removal rate (column), the first row lists the average percentage of downsampling zeros identified by mbImpute; the second row lists the average Pearson correlation between a downsampled matrix and the original matrix (on the log-scale) before imputation; the third row lists the average Pearson correlation (on the log-scale) after mbImpute is used. Each average calculated across the ten downsampling and is accompanied with an error margin, i.e., 1.96 times the standard error over the ten downsampling

ISSN: 1474-760X