install.packages( "lipidomeR" )
library( "lipidomeR" )

# Load the humanlipidome data set.

data( lipidomeR::humanlipidome )

# Transform the concentrations into log-10 scale.

humanlipidome$'Concentration_log10_umol_per_mL' <-
   log10( humanlipidome$'Concentration' )

# Enumerate the lipid names into values.

names.mapping <- map_lipid_names( x = humanlipidome$'Name' )

# Create the lipidomeR heatmap of lipid concentrations.

heatmap_lipidome(
   x = humanlipidome[ , c( "Name", "Concentration_log10_umol_per_mL" ) ],
   names.mapping = names.mapping,
   class.facet = "wrap",
   x.names = "Name",
   fill.limits =
       range(
           x = humanlipidome$"Concentration_log10_umol_per_mL",
           na.rm = TRUE
       ),
   fill.midpoint =
       sum(
           range(
               x = humanlipidome$"Concentration_log10_umol_per_mL",
               na.rm = TRUE
           )
       ) / 2,
   melt.value.name = "Concentration_umol_per_mL_log10",
   scales = "free"
)

# Load the cancerlipidome data set.

data( lipidomeR::cancerlipidome )

# Convert the data into wide format, where each lipid is one column and
# each sample is one row.

cancerlipidome.wide <-
   tidyr::pivot_wider(
       data = cancerlipidome,
       names_from = Lipid_Name,
       values_from = Lipid_Level
   )

# Inspect the data frame.
# View( cancerlipidome.wide )

# Create a mapping of the lipid names.

names.mapping <-
   map_lipid_names( x = unique( cancerlipidome$"Lipid_Name" ) )

# Compute the regression models.

result.limma <-
   compute_models_with_limma(
       x = cancerlipidome.wide,
       dependent.variables = names.mapping$"Name",
       independent.variables = c( "Group" )
   )

# Create the figure of all lipids and factors.

figure.output <-
   heatmap_lipidome_from_limma(
       x = result.limma$"model",
       names.mapping = names.mapping,
       axis.x.carbons = FALSE,
       class.facet = "row",
       plot.all = TRUE,
       plot.individual = FALSE,
       print.figure = TRUE,
       scales = "free",
       space = "free"
   )

# Create factor-specific figures.

figure.output <-
   heatmap_lipidome_from_limma(
       x = result.limma$"model",
       names.mapping = names.mapping,
       axis.x.carbons = FALSE,
       class.facet = "wrap",
       omit.class = "PA",
       plot.all = FALSE,
       plot.individual = TRUE,
       print.figure = FALSE,
       scales = "free",
       space = "free"
   )

# Print the figure of differences between cancer and benign tumors.

print( figure.output[[ "GroupCancer" ]] )

# Load the liverlipidome data set.

data( lipidomeR::liverlipidome )

# Convert the data into wide format, where each lipid is one column and
# each sample is one row.

liverlipidome.wide <-
   tidyr::pivot_wider(
       data = liverlipidome,
       names_from = Lipid_Name,
       values_from = Lipid_Level
   )

# Create a mapping of the lipid names.

names.mapping <-
   map_lipid_names( x = unique( liverlipidome$"Lipid_Name" ) )

# Compute the regression models.

result.limma <-
   compute_models_with_limma(
       x = liverlipidome.wide,
       dependent.variables = names.mapping$"Name",
       independent.variables = c( "Diagnosis" ),
       F.test = TRUE # Compute an F-test for a factor variable.
   )

# Compute the F-test.

result.limma <-
   compute_F_test_with_limma(
       x = result.limma,
       print.table = FALSE
   )

# Print a figure of the F-test.

figure.output <-
  heatmap_lipidome_from_limma(
      x = result.limma,
      names.mapping = names.mapping,
      F.test = TRUE,
      axis.x.carbons = FALSE,
      class.facet = "wrap",
      plot.all = FALSE,
      plot.individual = TRUE,
      scales = "free",
      space = "free"
  )

# Compute pairwise post-hoc comparisons between the factor levels for
# the dependent variables (i.e., lipids) with a significant F-test result.

result.limma <-
   compute_post_hoc_test_with_limma(
       x = result.limma,
       remap.level.names = TRUE
   )

# Print a figure of all post-hoc comparisons.

figure.output <-
    heatmap_lipidome_from_limma(
    x = result.limma$"result.post.hoc.test",
    names.mapping = names.mapping,
    axis.x.carbons = FALSE,
    plot.all = TRUE,
    plot.individual = FALSE,
    scales = "free",
    space = "free"
)

# Specify the contrasts of the post-hoc comparison that will be included
# in the figure.

contrasts.included <-
   c( "DiagnosisSteatosis", "DiagnosisNASH", "DiagnosisCirrhosis" )

# Get the omitted contrasts based on the above definition.

contrasts.omitted <-
   colnames( result.limma$"result.post.hoc.test"$"p.value" )[
       !(
           colnames( result.limma$"result.post.hoc.test"$"p.value" ) %in%
           contrasts.included
       )
   ]

# Find dependent variables (i.e., lipids) that have any significant
# difference.

has.any.significant <-
   apply(
       X =
           result.limma$"result.post.hoc.test"$"p.value"[
               ,
               contrasts.included
           ],
       MAR = 2,
       FUN = p.adjust,
       method = "BH"
   )

has.any.significant <-
   rownames(
       has.any.significant[
           apply(
               X = has.any.significant < 0.05,
               MAR = 1,
               FUN = any
           ),
       ]
   )

# Include in the figure only lipid classes that have at least four
# significant differences.

classes.included <-
   names(
       which(
           table(
               names.mapping[
                   make.names( has.any.significant ), "Class"
               ]
           ) > 3
       )
   )

classes.omitted <- unique( names.mapping$"Class" )
classes.omitted <-
   classes.omitted[ !( classes.omitted ) %in% classes.included ]

# Print a figure of the selected post-hoc-comparisons.

figure.output <-
   heatmap_lipidome_from_limma(
       x = result.limma$"result.post.hoc.test",
       names.mapping = names.mapping,
       axis.x.carbons = FALSE,
       omit.class = classes.omitted,
       omit.factor = contrasts.omitted,
       plot.all = TRUE,
       plot.individual = FALSE,
       scales = "free",
       space = "free"
   )

utils::sessionInfo()