Used rarefided or unrarefided feature-table

[suiweikang]

210 samples were sequencing in my research. And the number of sequencing reads of my sample was varied greatly, from 40,000 to 210,000. Also, the number of sequencing reads was rarefied to 40,000.
Should rarefied or unrarefied feature-table used in decontam?

[benjjneb]

By default decontam will normalize the data to proportions. And so rarefaction is not recommended, especially if negative controls are included which often have significantly lower read counts.

[suiweikang]

Thanks for the reply. Are both the frequency or prevalance method based on the proportion? Or is frequency based on proportions, while prevalence based on absolute read number?

[benjjneb]

The prevaaece method is based on presence/absence, so a proportion > 0.