Euclidean distance and CLR transformation

[MilesMatthews]

Hello, thank you for developing this amazing tool and making it available!

My Professor and I are interested in using Euclidean distances to calculate the beta diversity, and using central log ratio to transform the data. Is there a way we can do this for the generate_beta_trend_test_long function?

I’ve attached the code where we are using the generate_beta_trend_test_long function, with Bray-Curtis distance instead of Euclidean distance. Thank you so much!

5_MicrobiomeStat_Beta_Diversity copy.R.zip

[cafferychen777]

Hello Miles,

Thank you for your interest in MicrobiomeStat and for bringing up this important question about using Euclidean distance with CLR transformation in the generate_beta_trend_test_long function. I appreciate you sharing this with the community, as it will likely benefit other users as well.

While Euclidean distance isn't directly available as a default option in generate_beta_trend_test_long, we can implement it using a custom approach. Here's a solution that allows you to use Euclidean distance on CLR-transformed data within the MicrobiomeStat framework:

1. First, convert your phyloseq object to a MicrobiomeStat data object:

coral_data.obj <- mStat_convert_phyloseq_to_data_obj(phy.all)

2. Then, use the following function to perform CLR transformation and calculate Euclidean distance:

library(vegan)  # Used for calculating distance

clr_transform_and_calculate_distance <- function(data_obj) {
  # Extract feature table
  feature_tab <- data_obj$feature.tab
 
  # Manually implement CLR conversion
  clr_transform <- function(x) {
    # Add a small value to avoid log(0)
    x <- x + 1e-6
    # Calculate the geometric mean of each sample
    g_mean <- exp(rowMeans(log(x)))
    # CLR conversion
    log(x / g_mean)
  }
 
  # Apply CLR transformation
  clr_data <- t(clr_transform(t(feature_tab)))
 
  # Calculate Euclidean distance
  dist_matrix <- vegdist(t(clr_data), method = "euclidean")
 
  # Create dist_obj
  dist_obj <- list(
    CLR_Euclidean = as.matrix(dist_matrix)
  )
 
  return(dist_obj)
}

# Use the function
dist_obj <- clr_transform_and_calculate_distance(coral_data.obj)

3. Now you can use this dist_obj in the generate_beta_trend_test_long function:

all_times_analysis <- generate_beta_trend_test_long(
  data.obj = coral_data.obj,
  dist.obj = dist_obj,
  subject.var = "Colony.ID",
  time.var = "Collection.time.num",
  group.var = "Coral.species",
  adj.vars = c("Depth"),
  dist.name = "CLR_Euclidean"
)

# View the results
all_times_analysis$CLR_Euclidean

This approach allows you to use Euclidean distance on CLR-transformed data within the MicrobiomeStat framework. The dist.name argument is set to "CLR_Euclidean" to match the name we gave to the distance matrix in the dist_obj.

I hope this helps you and others in the community who might be interested in using Euclidean distance with CLR transformation. If you have any further questions or run into any issues implementing this solution, please don't hesitate to ask.

Thank you for contributing to the MicrobiomeStat community!

Best regards,
Chen Yang

[MilesMatthews]

Thank you so much! This will be really helpful for our research.