Skip to content

Commit

Permalink
Add data description
Browse files Browse the repository at this point in the history
  • Loading branch information
@fouodo
fouodo committed Nov 26, 2024
1 parent fe9bfbc commit 74df94e
Show file tree
Hide file tree
Showing 3 changed files with 493 additions and 128 deletions.
23 changes: 23 additions & 0 deletions R/multi_omics.R
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
#' Simulated multiomics data for 70 training participants and 23 testing participants,
#' each with an effect size of 20 on each layer. Each layer includes 50 participants for
#' training and 20 for testing. Participants do not perfectly overlap across layers.
#' The simulation is based on the R package \code{interSIM}.

#'
#' The dataset is a list containing training and testing data,
#' called \code{training} and \code{testing} respectively. Each data is a list
#' containing the following multi_omics at each layer.
#'
#' \itemize{
#' \item \code{methylation}: A \code{data.frame} containing the simulated methylation dataset.
#' \item \code{genexpr} : A \code{data.frame} containing the gene expression dataset.
#' \item \code{proteinexpr}: A \code{data.frame} containing the protein expression dataset.
#' \item \code{target}: A \code{data.frame} with two columns, containing patient IDs and values of target variable.
#' }
#'
#' @docType data
#' @keywords datasets
#' @name multi_omics
#' @usage data(multi_omics)
#' @format A list with training and testing data contaning methylation, gene expressions and protein expressions data.
"multi_omics"
74 changes: 36 additions & 38 deletions README.Rmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -63,10 +63,10 @@ Two types of data were simulated: training and testing datasets. Each consists o


```{r data_exam, include=TRUE, eval=TRUE}
data("entities")
data("multi_omics")
# This is a list containing two lists of data: training and test.
# Each sublist contains three entities.
str(object = entities, max.level = 2L)
# Each sublist contains three omics data.
str(object = multi_omics, max.level = 2L)
```

Variable selection, training and prediction are the main functionalities of `fuseMLR`. We can perform variable selection, train and fuse models for training studies, and predict new studies.
Expand All @@ -79,8 +79,8 @@ We need to set up training resources.
training <- createTraining(id = "training",
ind_col = "IDS",
target = "disease",
target_df = entities$training$target,
verbose = FALSE)
target_df = multi_omics$training$target,
verbose = TRUE)
print(training)
```

Expand All @@ -90,50 +90,56 @@ print(training)
# Create gene expression layer
createTrainLayer(training = training,
train_layer_id = "geneexpr",
train_data = entities$training$geneexpr,
train_data = multi_omics$training$geneexpr,
varsel_package = "Boruta",
varsel_fct = "Boruta",
varsel_param = list(num.trees = 1000L,
mtry = 3L,
probability = TRUE),
probability = TRUE,
na.action = "na.learn"),
lrner_package = "ranger",
lrn_fct = "ranger",
param_train_list = list(probability = TRUE,
mtry = 1L),
mtry = 1L,
na.action = "na.learn"),
param_pred_list = list(),
na_rm = TRUE)
na_rm = FALSE)
# Create gene protein abundance layer
createTrainLayer(training = training,
train_layer_id = "proteinexpr",
train_data = entities$training$proteinexpr,
train_data = multi_omics$training$proteinexpr,
varsel_package = "Boruta",
varsel_fct = "Boruta",
varsel_param = list(num.trees = 1000L,
mtry = 3L,
probability = TRUE),
probability = TRUE,
na.action = "na.learn"),
lrner_package = "ranger",
lrn_fct = "ranger",
param_train_list = list(probability = TRUE,
mtry = 1L),
mtry = 1L,
na.action = "na.learn"),
param_pred_list = list(),
na_rm = TRUE)
na_rm = FALSE)
# Create methylation layer
createTrainLayer(training = training,
train_layer_id = "methylation",
train_data = entities$training$proteinexpr,
train_data = multi_omics$training$proteinexpr,
varsel_package = "Boruta",
varsel_fct = "Boruta",
varsel_param = list(num.trees = 1000L,
mtry = 3L,
probability = TRUE),
probability = TRUE,
na.action = "na.learn"),
lrner_package = "ranger",
lrn_fct = "ranger",
param_train_list = list(probability = TRUE,
mtry = 1L),
mtry = 1L,
na.action = "na.learn"),
param_pred_list = list(),
na_rm = TRUE)
na_rm = FALSE)
```

- Also add a meta layer.
Expand All @@ -146,7 +152,7 @@ createTrainMetaLayer(training = training,
lrn_fct = "weightedMeanLearner",
param_train_list = list(),
param_pred_list = list(),
na_rm = FALSE)
na_action = "na.impute")
```

- An upset plot of the training data: Visualize patient overlap across layers.
Expand All @@ -162,8 +168,7 @@ Perform variable selection on our training resources
```{r varsel, include=TRUE, eval=TRUE}
# Variable selection
set.seed(5467)
var_sel_res <- varSelection(training = training,
verbose = FALSE)
var_sel_res <- varSelection(training = training)
print(var_sel_res)
```

Expand All @@ -178,10 +183,8 @@ set.seed(5462)
training <- fusemlr(training = training,
use_var_sel = TRUE,
resampling_method = NULL,
resampling_arg = list(y = entities$training$target$disease,
k = 10L),
impute = TRUE,
verbose = FALSE)
resampling_arg = list(y = multi_omics$training$target$disease,
k = 10L))
print(training)
# See also summary(training)
Expand All @@ -206,17 +209,17 @@ testing <- createTesting(id = "testing",
# Create gene expression layer
createTestLayer(testing = testing,
test_layer_id = "geneexpr",
test_data = entities$testing$geneexpr)
test_data = multi_omics$testing$geneexpr)
# Create gene protein abundance layer
createTestLayer(testing = testing,
test_layer_id = "proteinexpr",
test_data = entities$testing$proteinexpr)
test_data = multi_omics$testing$proteinexpr)
# Create methylation layer
createTestLayer(testing = testing,
test_layer_id = "methylation",
test_data = entities$testing$proteinexpr)
test_data = multi_omics$testing$proteinexpr)
```

- An upset plot of the training data: Visualize patient overlap across layers.
Expand All @@ -237,7 +240,7 @@ print(predictions)
```{r performance_all, include=TRUE, eval=TRUE}
pred_values <- predictions$predicted_values
actual_pred <- merge(x = pred_values,
y = entities$testing$target,
y = multi_omics$testing$target,
by = "IDS",
all.y = TRUE)
x <- as.integer(actual_pred$disease == 2L)
Expand Down Expand Up @@ -275,12 +278,10 @@ We distinguish common supervised learning arguments from method specific argumen
The interface approach leverages the arguments in ```createTrainLayer()``` to map the argument names of the original learning function. In the example below, the gene expression layer is re-created using the ```svm``` (Support Vector Machine) function from the ```e1071``` package as the learner. A discrepancy arises in the argument names of the ```predict.svm``` function, which uses ```object``` and ```newdata```.

```{r interface, include=TRUE, eval=TRUE}
# Remove the current gene expression layer from training
removeLayer(training = training, layer_id = "geneexpr")
# Re-create the gene expression layer with support vector machine as learner.
createTrainLayer(training = training,
train_layer_id = "geneexpr",
train_data = entities$training$geneexpr,
train_data = multi_omics$training$geneexpr,
varsel_package = "Boruta",
varsel_fct = "Boruta",
varsel_param = list(num.trees = 1000L,
Expand All @@ -304,12 +305,10 @@ createTrainLayer(training = training,
)
# Variable selection
set.seed(5467)
var_sel_res <- varSelection(training = training,
verbose = FALSE)
var_sel_res <- varSelection(training = training)
set.seed(5462)
training <- fusemlr(training = training,
use_var_sel = TRUE,
verbose = FALSE)
use_var_sel = TRUE)
print(training)
```
Expand Down Expand Up @@ -355,11 +354,10 @@ createTrainMetaLayer(training = training,
lrner_package = NULL,
lrn_fct = "mylasso",
param_train_list = list(nlambda = 100L),
na_rm = TRUE)
na_action = "na.impute")
set.seed(5462)
training <- fusemlr(training = training,
use_var_sel = TRUE,
verbose = FALSE)
use_var_sel = TRUE)
print(training)
```

Expand Down
Loading

0 comments on commit 74df94e

Please sign in to comment.