Paired sample setting and Q value

Downstream analysis and statistics MaAsLin
1

MaAsLin2 1.12.0

Hi,

I have two problems in MaAsLin2. My first problem is the paired sample setting. My original sample ID was something like:
1, 2, 3, 4, 5
1T, 2T, 3T, 4T, 5T

Here, I only have one factor and two variables (treatment condition; Yes and no). Then I run the MaAsLin2 and I found that MaAsLin2 does not run the ‘paired’ test. In fact, it is running something like an ‘unpaired’ test. After I found that, I tried to set my sample ID like:

1, 2, 3, 4, 5
1, 2, 3, 4, 5

I have not changed the factor and two variables. However, the system was shown as ‘repeated row name’. In this condition, how can I set the sample ID, metadata and MaAsLin2 to let MaAsLin2 run a ‘paired’ test?

Another question is for the Q value. For small sample-size projects (e.g., N<15), the Q value is not reliable and I want MaAsLin2 only consider the P value. How can I set it in MaAsLin2? I know I can set max_significance to 0.6 or more.

Best wishes,
Shuyuan

2

Hi @Shuyuan_Zhang,

Under the hood Maaslin2 is running a linear model and as such does not have a “paired” test. However if you are interested in doing something like a paired test you could run a mixed linear model in Maaslin2 with a random intercept for each sample ID (1,2,3,4,5).

Maaslin2 does report the p-values associated with fixed effects. You can use the results data.frame and/or all_results.tsv file to inspect those p-values and determine which are of interest to you.

Thanks,
Jacob Nearing

3

Dear Jacob,

Sorry for my late reply. Did you mean to add command ‘random_effects’ in the MaAsLin 2 settings? To run the randon_effects, does it means adding another column (e.g., paired list) in the metadata file and then the data in paired list should be something like ‘1,1,2,2,3,3,4,4,5,5’?

Best wishes,
Shuyuan

4

Hi @Shuyuan_Zhang,

Sorry for the late response… but yes I would suggest adding random effects using the random_effects parameter in the Maaslin2 command. To run the random effects you would want to include a metadata column that indicates which samples are paired.

Cheers,
Jacob Nearing