TimeOmics recommended minimum number of time points

Hi there,

I was successful in using the MixOmics platform to integrate and analyze my omic datasets:

  • Sphingolipids (30 sphingolipids),
  • Microbiome (60 taxa)),
  • Inflammation (8 cytokines)

I used the TSS method to normalize my microbiome data and filtered out low abundance taxa.
For my sphingolipid and cytokine datasets I performed a median normalization and autoscaling.
Finally I used the multi-block sPLS-DA method as outlined in the mixOmics tutorial.

The results were quite neat and informative, but I have a few more datasets that I’d like to integrate. The datasets are from earlier collection timepoints.

We have:
2 sphingolipid datasets (corresponding to sampling at 12 and 21 months)
4 microbiome datasets (corresponding to sampling at 0, 6, and 16 and 21 months)
2 inflammation datasets (corresponding to sampling at 12 and 21 months)

From reading the paper, I noticed that having mismatching timepoints shouldn’t be problem for timeOmics, however, I only have 2 datasets for my inflammatory and sphingolipid data. Would it still be fruitful to analyze these data longitudinally using the timeOmics package? Any advice would be helpful.

hi @mak0130

I am glad to know that mixOmics is working for you. In the Bodein et al paper (Discussion), we advise to have about 5 time points. In your case that won’t work also because you would need to extrapolate the times 0 and 6 (i.e outside the time range of 12-21).

Have you looked at the multilevel approach? (there are a few posts about it I think, explaining how you should extract the withinVariation() first before inputting these data in a block.splsda().