ANCOM cutoff determination

Hello QIIME developers/experts,

I have run ANCOM to compare genera distribution across different batches(k = 3). As seen in the attached qzv, g_Chromobacterium has a significantly different distribution. I did a quick visualization with STAMP, and found that the genus is especially abundant in Batch3.

Both g__Barnisiella and g_[Ruminococcus]torques_group appear to have quite high W statistics but are not said to be statistically significant. A quick STAMP barplot visualization would suggest that both are (visually) differentially abundant in Batch1 samples.

I have read through ANCOM discussions in the forum. I also went through the paper and it's supplementary material, but forgive me; I am not from a strong statistical background, and I cannot wrap my head around how W cutoff was determined.

The genera table contains 583 observations. I understand that W stands for the number of subhypotheses being rejected. A score of 399 can be translated as the log ratio for Chromobacterium unequal to log ratio of other 399 genera. From the plot, W cutoff would be in between 399 (g__Chromobacterium) and 336(g__Barnesiella).

Can you tell me how QIIME2/ANCOM compute this cutoff, in layman terms hopefully?

GENUS_ancom_batch.qzv (394.0 KB)

2 Likes

Hi @ZarulSaurus, that is a very good question - something that may be more appropriate ask either the first author or the corresponding author of the ANCOM paper.

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.