search

theislab / scCODA

A Bayesian model for compositional single-cell data analysis
BSD 3-Clause "New" or "Revised" License
141 stars 23 forks source link

Usage without replicates ? #90

Closed divyanshusrivastava closed 4 months ago

divyanshusrivastava commented 4 months ago

Hi. I tried using the package on my data, where I have control and case conditions with only one replicate each. The model returned no significant changes, and I was wondering if this is because of lack of replicates. I followed this tutorial.

Can you comment on how to use the package when one has no replicates ? Thanks

johannesostner commented 4 months ago

Hi @divyanshusrivastava! It's very likely that the lack of replicates is one reason why you don't get any significant results. As with any other statistical model or test, scCODA's uncertainty depends on the number of replicates. If there's very few, only very strong differences wil be reliably detected.

divyanshusrivastava commented 4 months ago

Thanks for the response :) Closing this issue