Alpha Diversity question

[RachBioHaz]

Hi, I was looking over the alpha diversity example and you state that if rarefying isn't used as a normalisation step, precalculated alpha diversity indices can be passed to the alpha.obj parameter in the alpha diversity analysis functions. What format does this need to be in and is there an example I can follow that does not use rarefying? Thanks

[cafferychen777]

Hi there,

Thank you for reaching out with your question about using precalculated alpha diversity indices with our R package's alpha diversity analysis functions.

If you choose not to rarefy your data as a normalization step, you can indeed pass precalculated alpha diversity indices to the alpha.obj parameter.

Here is an example of how you can calculate the alpha diversity indices and then pass them to the alpha.obj parameter:

# Calculate alpha diversity indices
alpha.obj <- mStat_calculate_alpha_diversity(
  x = data.obj$feature.tab,
  alpha.name = c("shannon", "simpson", "observed_species", "chao1", "ace", "pielou")
)

# Now, you can pass this `alpha.obj` directly to the alpha diversity analysis function
# For example, if you have a function `alpha_diversity_analysis`, you would do:
result <- mStat_alpha_xxxx(alpha.obj = alpha.obj, ...)

# Replace '...' with any other arguments your function requires.

Please make sure that the column names of alpha.obj match the expected diversity indices that the analysis function is looking for. If you need further assistance or have any more questions, feel free to ask.

Best regards,

Chen YANG

[adahal123]

Following up on this...
I created an alpha.obj with shannon and observed_species. It works great for generate_alpha_trend_test_long and generate_alpha_volatility_test_long.

However, when using generate_alpha_test_long, it still defaults to rarefy. There is no difference between using it and not using it.

alpha_test <- generate_alpha_test_long(
data.obj = ps.obj,
alpha.obj = alpha.obj,
alpha.name = c("shannon","observed_species"),
time.var = "visit",
t0.level = unique(ps.obj$meta.dat$visit)[1],
ts.levels = unique(ps.obj$meta.dat$visit)[-1],
group.var = "race",
adj.vars = "sex"
)

...
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.
Diversity analysis needs rarefaction! Call 'mStat_rarefy_data' to rarefy the data!
No depth specified, using the smallest column sum: 490408
Data has been subsetted based on the provided samIDs.

Warning messages:
1: In vegan::rrarefy(t(otu_tab), sample = depth) :
function should be used for observed counts, but smallest count is 11
2: In vegan::rrarefy(t(otu_tab), sample = depth) :
function should be used for observed counts, but smallest count is 9
3: In vegan::rrarefy(t(otu_tab), sample = depth) :
function should be used for observed counts, but smallest count is 11

What am I doing wrong?

[adahal123]

This is the new error with the generate_alpha_test_long it seems after you have updated it:

Code:
alpha_test <- generate_alpha_test_long(
data.obj = ps.obj,
alpha.obj = alpha.obj,
alpha.name = c("shannon","observed_species"),
time.var = "visit",
t0.level = unique(ps.obj$meta.dat$visit)[1],
ts.levels = unique(ps.obj$meta.dat$visit)[-1],
group.var = "race",
adj.vars = "sex"
)

Result:
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.
Data has been subsetted based on the provided condition.
Updated metadata to match the subsetted data.
The following samples were excluded:
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.
Error in names(test.list) <- alpha.name :
'names' attribute [2] must be the same length as the vector [1]

As I had mentioned, this alpha.obj has worked in other longitudinal alpha function tests.

[cafferychen777]

Hi @adahal123,

Thank you for reporting this issue with the generate_alpha_test_long function. I took a closer look and it appears the error may be occurring because the alpha diversity metrics specified in alpha.name when creating the alpha.obj object do not match the metrics specified in alpha.name when calling generate_alpha_test_long.

The error message suggests the names attribute of the test.list object has length 1 while the alpha.name vector has length 2. This likely means the alpha.obj only contains a single diversity metric (either Shannon or observed species) while you are attempting to analyze both metrics by passing c("shannon","observed_species") to the alpha.name parameter in generate_alpha_test_long.

Can you please double check that the alpha metrics stored in alpha.obj match the ones you specify in the generate_alpha_test_long call? The alpha.name arguments need to be consistent between the two.

Let me know if you have any other questions! I appreciate you taking the time to test out the package and provide this helpful feedback.

Best regards,
Caffery

[cafferychen777]

Hi @adahal123,

I've fixed the issue in the construct_formula function where only the first variable in adj.vars was being included in the analysis when using multiple covariates. The updated package will be available for download in a few hours. Thank you for bringing this to my attention!

Best,
Chen YANG

[adahal123]

Thanks, multiple adj.vars works great now.

As for the generate_alpha_test_long, the metrics "shannon" and "observed_species" are in the alpha.obj. The same alpha.obj and alpha.name = c("shannon", "observed_species") works for all other alpha diversity measure except for that one. For example:

DOES NOT WORK

alpha_test <- generate_alpha_test_long(
  data.obj = ps.obj,
  alpha.obj = alpha.obj,
  alpha.name = c("shannon", "observed_species"),
  time.var = "visit",
  t0.level = 1,
  ts.levels = c("2", "3"),
  group.var = "race",
  adj.vars ="sex"
)

Results:
Rule 1 passed: data.obj is a list.
Rule 2 passed: meta.dat is a data.frame.
Rule 3 passed: The row names of feature.tab match the row names of feature.ann.
Rule 4 passed: The column names of feature.tab match the row names of meta.dat.
Rule 5 passed: feature.tab is a matrix.
Rule 6 passed: feature.ann is a matrix.
Please note: The data components should follow base R data.frame and matrix structures, not phyloseq's formal class.
Validation passed.
Note: Passing validation does not guarantee the absence of all data issues. Further data exploration may be needed.
Data has been subsetted based on the provided condition.
Updated metadata to match the subsetted data.
The following samples were excluded: [SAMPLES from visit 2 and visit 3]
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.
Data has been subsetted based on the provided condition.
Updated metadata to match the subsetted data.
The following samples were excluded:
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.
Error in names(test.list) <- alpha.name :
'names' attribute [2] must be the same length as the vector [1]

WORKS

alpha_test_change <- generate_alpha_change_test_long(
  data.obj = ps.obj,
  alpha.obj = alpha.obj,
  alpha.name = c("shannon", "observed_species"),
  time.var = "visit",
  t0.level = 1,
  ts.levels = c("2", "3"),
  subject.var = "pid",
  group.var = "race",
  adj.vars = "sex",
  alpha.change.func = "log fold change"
)

Results:
Data has been subsetted based on the provided samIDs.
Updated metadata to match the subsetted data.
The following samples were excluded: [SAMPLES from visit 3]
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.
Data has been subsetted based on the provided samIDs.
Updated metadata to match the subsetted data.
The following samples were excluded: [SAMPLES from visit 2]
Updated feature table to match the subsetted data.
Updated feature annotation to match the subsetted data.
Data subsetting complete. Returning updated data object.

alpha_volatility <- generate_alpha_volatility_test_long(
  data.obj = ps.obj,
  alpha.obj = alpha.obj,
  alpha.name = c("shannon","observed_species"),
  time.var = "visit",
  subject.var = "pid",
  group.var = "race",
  adj.vars = "sex"
)

Result:
The volatility calculation in generate_alpha_volatility_test_long relies on a numeric time variable.
Please check that your time variable is coded as numeric.
If the time variable is not numeric, it may cause issues in computing the results of the volatility test.
You can ensure the time variable is numeric by mutating it in the metadata.
Alpha diversity has been adjusted for the following covariates: sex.
Alpha diversity has been adjusted for the following covariates: sex.

[cafferychen777]

I've resolved the issue with the generate_alpha_test_long function not working with the metrics "shannon" and "observed_species" in alpha.obj. The updated function should now work correctly. However, it's important to ensure that t0.level and ts.levels are consistent with the data type of the time.var in data.obj$meta.dat. They should be numeric if time.var is numeric, and character if time.var is character. Here's an example of how to use the updated function:

alpha_test <- generate_alpha_test_long(
  data.obj = ps.obj,
  alpha.obj = alpha.obj,
  alpha.name = c("shannon", "observed_species"),
  time.var = "visit",
  t0.level = 1,  # Use numeric if "visit" is numeric in ps.obj$meta.dat
  ts.levels = c(2, 3),  # Use numeric if "visit" is numeric in ps.obj$meta.dat
  group.var = "race",
  adj.vars ="sex"
)

Please make sure to match the data type of t0.level and ts.levels with the data type of time.var in your dataset. If you encounter any further issues or have additional questions, feel free to reach out.