Paired samples collected on the same day - best way to analyze them?

Hi Everyone,

I have bronchial brushings and bronchoalveolar lavage (BAL) samples collected from patients on the same day. In other words, I have 64 samples (32 from each sample type) from 32 different subjects. I’d like to see if these samples differ from each other. Can I simply perform diversity analyses using this approach: brushings vs. bronchoalveolar lavage (as independent groups)?
Is there a way of running a paired analysis? I know that there is the qiime2-longitudinal plugin, but I’m not sure I can apply it here, as the paired samples were collected on the same day.

I’d appreciate any advice here.

Thanks,
FS

Yes, q2-longitudinal is definitely applicable here! Since you have paired samples, you can use the pairwise-distances and pairwise-differences methods to compare these pairs — these methods are appropriate for any kind of paired samples, not just temporal pairs. See the q2-longitudinal tutorial on qiime2.org for more details.

That works too!

Good luck!

1 Like

Dr. Bokulich,

Thanks very much!

Really appreciate the amazing support we have here in the Qiime2 forum.

Thanks,
FS

2 Likes

Dr. Bokulich,

I’m trying to use the pairwise-distances and pairwise-differences methods, but I ran into some issues.

My data is organized like this:
All Brushes samples (Pair 0)
All BAL samples (Pair 1)
Type indicates whether a sample is BAL or Brush.
Time of collection between paired samples is the same, as those were collected during the same bronchoscopy.

With the qiime longitudinal pairwise-differences method, I was able to get an output (which seems correct: it shows: 64 samples and Pair1 - Pair 0), but only if I do not include –p-group-column (as this parameter is optional).

The full command is described below:

qiime longitudinal pairwise-differences
–m-metadata-file Metadata.txt
–m-metadata-file shannon_vector.qza
–p-metric shannon
–p-state-column Pair
–p-state-1 0
–p-state-2 1
–p-individual-id-column SUBJECT_ID
–p-replicate-handling random
–o-visualization shannon-pairwise-differences.qzv

However, for the the pairwise-distances method, the –p-group-column is required, and when I try to use:

qiime longitudinal pairwise-distances
–i-distance-matrix bray_curtis_distance_matrix.qza
–m-metadata-file Metadata.txt
–p-state-column Pair
–p-state-1 0
–p-state-2 1
–p-individual-id-column SUBJECT_ID
–p-replicate-handling random
–o-visualization BC-fragment-pairwise-distances.qzv

There was a problem with the command:
(1/1) Missing option “–p-group-column”.

Running the same command with this option also results in other error:

qiime longitudinal pairwise-distances
–i-distance-matrix bray_curtis_distance_matrix.qza
–m-metadata-file Metadata_Oct2019-BIDC_BAL_Brushes-COMPLETE-with-clinical-data1.txt
–p-group-column Type
–p-state-column Pair
–p-state-1 0
–p-state-2 1
–p-individual-id-column SUBJECT_ID
–p-replicate-handling random
–o-visualization BC-fragment-pairwise-distances.qzv

Plugin error from longitudinal:

State 0.0 is not represented by any members of BAL group in metadata. Consider using a different group_column or state value.

Currently, I do not know how to resolve this last issue, all all my samples were from one timepoint.
How can I solve this issue? Is there a way of disabling --p-group-column for the qiime longitudinal pairwise-distances method?

Thanks in advance,
FS

Yep, please see some discussion here:

Since BAL is one of the states you are comparing, it should not also be used as the group column.

The general goal with this test is that you may have a set of paired samples in two different treatment groups, and this test will determine if one group experiences greater divergence between states (e.g., pre/post treatment) than the other. This works whether the pairs are of patients’ samples collected at two different times, or in your case two different collection sites. So this would apply to your case if you are collecting BAL and brush samples from two different groups — e.g., one group is receiving some treatment.

Another option is to run PCoA on this sample set, and then input the PCoA data to pairwise-differences. This will determine whether you see a directional effect on beta diversity.

1 Like

Hi Dr. Bokulich,

Thanks very much!

FS

1 Like

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