How is the data format for QCs in Time-Series Only?

How can I put Quality Control samples in the input data?
I’m trying to use the Time Series Only, in the Statistical Analysis [metadata table]. I tried to use the metadata file with the column “time” with a number different from the samples and the column “subject” with QC, but the error message appears: “ErrorFailed meta-data integrity check. Possible causes of error (last one being the most relevant): Make sure to have equal replicates for each time points: Found min: 6 max: 8 Maybe specify study design as Multiple factors / covariates”.

Hi Alessandra,

The ‘time series’ input is meant for experimental designs where you have measured the same subjects multiple times (for example, performing metabolomics on blood samples drawn from the same patients at multiple time points). The statistical method has quite strict requirements, for example no missing data, balanced design, etc.

MetaboAnalyst has some checks to make sure your data fits the required design. The error says you must have equal replicates for each time point: if you had, for example, 10 patients and you measured each for each time point, then the dataset would have equal # replicates (10) at each time point.

It sounds like your data does not fit this scenario. I suggest exploring the ‘Multiple factors/ covariates’ design as MetaboAnalyst suggested because it is much more flexible and can accommodate more types of experimental designs.


Hi Jess,

Thanks for answering.

I understand that is required the same number of samples for each subject, but the problem is that I need to load the quality control group (QC), which is a pool of samples. Do you know how to do this?


I’m not sure, @jeff.xia do you know how to format the data in this case?

In MetaboAnalyst, the key purposes of using QC:

  1. PCA overview;
  2. QC based data filtering

Since this is not related to actual data analysis, you can remove one factor, and use Statistical Analysis [one factor] for QC assurance. Then download the QC filtered data, add back the factor and upload to meta-data module

1 Like

Hi @jeff.xia and @jess.ewald,

I’m replying to this forum as its somewhat similar of an issue to my own, and I’m hoping you’d have an idea of what to do.
I am hoping to do a time series analysis on fish mucus protein data I collected. The problem I am facing is that I have data from 4 time points, with fish sampled in a pooled manner (15 fish in each tank; 5 fish pooled = one sample), from three different treatments (control, low, high) each with three replicates, making 9 samples per time point per treatment. There was no way to know which ‘subjects’ were sampled each time as it is a pool, where in your example you know with the human blood samples which individual you sampled at each time point. The pools of fish I sampled from each of the treatments at each time point thus should not be given the same ‘subject ID’ at each time point as outlined in the reference excel sheet.
Attached is my current meta data sheet, however as I said assigning the subject ID in this manner would likely be incorrect… is there a way to get around this issue and not treat the subjects as the same individual at each time point?
meta_data_mucusproteins.csv (1.4 KB)

We are unable to provide project-specific consulting, but perhaps someone in the user community can help you if they have similar experience to your dataset.

One general suggestion is to learn more about the “Linear Models with Covariate Adjustment” tool, in the multi-factor metadata table module (Multiple factors/covariates study design). It is very flexible and can accommodate almost any experimental design.