Hello guys,
May I please ask for help with qiime longtigudinal feature volatility analysis?
My metadata file contains 2 columns: individual-id-column called sampleid, and state-column called group. The format has passed the check with Keemei. But whenever I run the code, there has been the error indicating the "sampleid column is not a column in the metadata", which is definitely not the case. I pasted my code here, also a screenshot of the error.
I would really appreciate any feedback from you guys.
Thank you so much for your time and help!
Hello!
Longitudinal analyses assume that you have the same individuals / subjects sampled multiple times. So you need to indicate a corresponding column that includes such information, while in your current command you provide an index column.
Hello Timur,
Thank you so so much for your quick response! Following your point of adding a column to specify the individuals, if I add another column named "subjects" for instance, should I also add that information in the command? If so, where and what should I put in the command?
I really appreciate your help, TImur!
Hello Timur,
Problem solved! Thank you so much for your helpful information!
Two quick followup questions:
I' m not sure how to interpret the most important features here. I also did a linear mixed effects analysis to check the most important changes. How should I correlate both results and interpret them properly in the manuscript? Besides, there are only 10 features being reported. Can I have more? Can't find any bottons on the "QIIME2 View" page
Can I only focus on certain patterns in the longitudinal volatility analysis? For instance, I have 4 timepoints in total. Can I find the pattern of the features that rise in the first 3 timepoints but drop in the very last timepoint? If so, what command should I add?
Thank you very much again, Timur! I really appreciate it! A screenshot of the current QIIME2 View page is pasted here FYI
Hello!
I'm glad that you already made some progress!
The most important features here are the features that either change their relative abundances over time or are strong predictors of one of the time-points.
Not sure if I can answer here. I would start by looking for important features that predict changes you found by LME.
Redoing everything at the ASV level will probably yield more important features. It may happen that there are some ASVs that were assigned to the same taxa but demonstrate different patterns over time. But I can't say for sure without testing it.
Also not sure that you can be so specific within qiime2. But if you have only 10 features that are important you always can inspect them one by one. Otherwise some smart filtering/scripting in Excel / R / Python may help you to identify such features.
