Changing the order of factors in TSE

[RachBioHaz]

Hi, I'm new to mia and I wanted to know how factor level order can be changed. Is it possible to set the factor levels to a specific order, so that my reference group is always first?
Cheers
Rach

[antagomir]

Do you mean in colData(tse)?

[TuomasBorman]

Hello @RachBioHaz , thanks for your question! Hopefully you find this conversation. I moved it here so that we have only one discussion forum open.

[RachBioHaz]

Thanks @TuomasBorman

[RachBioHaz]

Hi @antagomir,
yes, the colData(tse). I'm just learning to use mia and am trying to get my head around manipulating the data in it.
I was struggling getting my factors to work, not sure why. Ended up doing it this way:

set factor for $variable

data.obj$meta.dat$variable <- factor(
as.factor(data.obj$meta.dat$variable),
levels = c("group_a", "group_b", "group_c", "group_d")
)

Let me know if there's a better way, especially when converting a phyloseq object (with existing factors) to tse.
Cheers
Rachele

[TuomasBorman]

Your idea is correct. colData() is just data.frame. You can modify the columns of it similarly to base R. tse$SampleType does the same thing as colData(tse)$SampleType.

library(mia)
data("GlobalPatterns")
tse <- GlobalPatterns

# Check colData
colData(tse)

# Specify order of SampleType
tse$SampleType <- factor(
    tse$SampleType,
    levels = c("Skin", "Soil", "Tongue", "Feces", "Mock Ocean Sediment (estuary)", "Freshwater (creek)", "Freshwater", "Mock")
    )

As you are using syntax data.obj$meta.dat$variable, you are first specifying "meta.dat" column from colData(data.obj). The column seems to be a list, and you retrieve the "variable" element from it.

Columns of colData() can be lists but usually that is not the optimal way to do it. Instead, columns should be vectors (such as "SampleType" column in "GlobalPatterns" data).

To debug this:

1. Did you create TreeSE object manually or did you use convertFromPhyloseq?
   • Is the sample metadata column in phyloseq object also a list?
2. Can you print here your data.obj?
3. Can you print here data.obj$meta.dat$variable and colData(data.obj)?

[RachBioHaz]

Hi,
I did use the convertFromPhyloseq function but for some reason, the factors did not carry across when I first tried it. I tried it again just now with the original phyloseq object and it worked. I had obviously made some error in my prior attempt, as it's working fine now.

Thanks for your help.