ANCOMBC2 code

[Lobna-H]

I am trying to perform differential abundance for microbial community composition by ANCOMBC2 using R.
I have 3 samples from 4 dependent groups (group 1, group 2, group 3, group 4) each group contains 3 samples. group1 is the beginning point (timepoint0), group 2 is group 1 after treatment 1 for a period of time (timepoint1), group 3 is group 2 after treatment 2 for a period of time (timepoint3), group 4 is group 3 after treatment 3 for a period of time (timepoint4). group 1 is the reference group.

Here my script

output <- ancombc2(data = pseq5, assay_name = "counts", tax_level = "Phylum",
fix_formula = " groups + Timepoint",
rand_formula = "(Timepoint | subject)",
p_adj_method = "holm",
prv_cut = 0.10, lib_cut = 1000, s0_perc = 0.05,
group = "groups", struc_zero = TRUE, neg_lb = TRUE,
alpha = 0.05, n_cl = 3, verbose = TRUE,
global = TRUE, pairwise = TRUE, dunnet = TRUE, trend = TRUE,
iter_control = list(tol = 1e-2, max_iter = 20,
verbose = TRUE),
em_control = list(tol = 1e-5, max_iter = 100),
lme_control = lme4::lmerControl(),
mdfdr_control = list(fwer_ctrl_method = "holm", B = 100),
trend_control = list(contrast = list(matrix(c(1, 0, 0,
-1, 1, 0,
0, -1, 1),
nrow = 3,
byrow = TRUE)),
node = list(3),
solver = "ECOS",
B = 100))

I got the error

Obtaining initial estimates ...
Error: Estimation failed for the following covariates:
Timepointt1, Timepointt2, Timepointt3
Please ensure that these covariates do not have missing values and check for multicollinearity before re-estimating the model
In addition: Warning message:
Small sample size detected for the following group(s):
group1, group2, group3, group4
Variance estimation would be unstable when the sample size is < 5 per group

Could you please help solve this error

[FrederickHuangLin]

Hello @Lobna-H,

If I've grasped your code correctly, it appears that your timepoint and group variables contain identical information, potentially leading to a multicollinearity problem. To investigate whether this issue exists, you can perform a simple linear regression lm using the formula groups + Timepoint with your microbial abundance data.

Best regards,
Huang

[simone-anza]

Hi @FrederickHuangLin

thank you for the reply, I am not sure about this approach and I am having the same issue of understanding which variable is multicollinear for my ancombc model for gut microbiome analyses. Let's say I want to check for multicollinearity of varA and varB, what is my response variable in my lm considering that microbial abundance data is compositional? Thank you for this tool!

Best,
Simone