Manual: abundance table generation with R

You can use the R package samovaR to perform data augmentation

Get your data

Your own data

# from .csv abundance table
data <- read_samovar("path/to/abundance_table.csv")
# from abundance table
data <- table2samovar(abundance_table)
# from phyloseq
data <- phyloseq2samovar(phyloseq_data)

Data from GMrepo.

If you want to use your abundance file, skip this stage

data <- GMrepo_type2data(mesh_ids)

#or to use with run list:
data <- GMrepo_run2data(runs)

Output - SAMOVAR object

Preprocessing

On this stage, you can split your data into groups and filter it using metadata

data %>% samovar_preprocess()

# Similar to:
## samovar_raw %>%
##   teatree_trim() %>%
##   tealeaves_pack() %>%
##   teabag_brew() %>%
##   concotion_pour()

This stage include:

• data filtration
• normalization
• database builidng

Boil the samovar to get new data

new_data <- samovar_data %>% samovar_boil(N = 100)

Output - SAMOVAR object

You also can use our web application

Comparison of GMrepo cohorts (moving pictures)

Generate samovar

datasets <- list()

generate_dataset <- function(meshID) {
  message(meshID)
  data <- meshID %>% 
    GMrepo_type2data(number_to_process = 1500)
  
  build <- data %>% 
    samovar_preprocess(min_cluster_size = 20, 
                       treshhold_amount = 10^(-5), 
                       treshhold_species = 5,
                       treshhold_samples = 5)
  
  datasets[[meshID]] <- build
  
  new_data <- build %>% 
    samovar_boil(N = 200)
  
  result <- rbind(
    data$data %>% 
      rownames_to_column("sp") %>% 
      pivot_longer(-1, names_to = "run") %>% 
      mutate(type = "initial", meshID = meshID),
    new_data$data %>% 
      rownames_to_column("sp") %>% 
      pivot_longer(-1, names_to = "run") %>% 
      mutate(type = "generated", meshID = meshID)
  )
  
  return(result)
}

meshIDs <- c(
  "D006262",
  "D003967",
  "D000855",
  "D003863",
  "D015179"
)

results <- tibble()
for(meshID in meshIDs){
  tmp <- generate_dataset(meshID)
  results <- rbind(results, tmp)
}

Vizualize samovar

results_ord <- function(type) {
  wide <- results %>% 
    subset(type == type) %>% 
    mutate(fn = paste(run, 
                      #type,
                      meshID)) %>% 
    pivot_wider(id_cols = sp, values_from = value, values_fill = 0, names_from = fn) %>% 
    column_to_rownames("sp")  %>% 
    subset(apply(., 1, sum) > 1)

  wide_dist <- wide %>% 
    t %>% 
    subset(apply(., 1, sum) > .1) %>% 
    vegan::vegdist() %>% 
    ecodist::pco()
  
  wide_dist$vectors %>% 
    as.data.frame() %>% 
    select(X1, X2) %>% 
    rownames_to_column("tp") %>% 
    mutate(disease = tp %>% str_remove(".* ")) %>% 
    ggplot() +
    theme_bw() +
    geom_point(aes(color = disease, X1, X2)) + 
    xlab("PC1") + ylab("PC2") +
    scale_color_brewer("",palette = "Set2")
}