Pseudocounts for compatibility with ordination

Downstream analysis and statistics MaAsLin
1

I would like to use the transformations as performed by maaslin3 in other downstream visualizations and analyses, so that the data are only processed once rather than re-processing with different tools and in slightly different ways for each analysis.

Would you consider implementing a pseudo-count option for maaslin3 so that the output “data_transformed.tsv” does not contain NA values? The NA values do not work well with PCA, for example.

2

Hi,

If you set the transform as "PLOG" when you run it, do you get the pseudo-count imputed data you’re looking for? The NAs are there in the default "LOG" transformation to tell the linear modeling system to skip the abundances that were 0.

Will

3

Ah - yes, that works if I also add the following options

evaluate_only = ‘abundance’,
zero_threshold = -1,

1 Like