BioM2 {BioM2}

Biologically Explainable Machine Learning Framework

Description

Biologically Explainable Machine Learning Framework

Usage

BioM2(
  TrainData = NULL,
  TestData = NULL,
  pathlistDB = NULL,
  FeatureAnno = NULL,
  resampling = NULL,
  nfolds = 5,
  classifier = "liblinear",
  predMode = "probability",
  PathwaySizeUp = 200,
  PathwaySizeDown = 20,
  MinfeatureNum_pathways = 10,
  Add_UnMapped = TRUE,
  Unmapped_num = 300,
  Add_FeartureSelection_Method = "wilcox.test",
  Inner_CV = TRUE,
  inner_folds = 10,
  Stage1_FeartureSelection_Method = "cor",
  cutoff = 0.3,
  Stage2_FeartureSelection_Method = "RemoveHighcor",
  cutoff2 = 0.95,
  classifier2 = NULL,
  target = "predict",
  p.adjust.method = "fdr",
  save_pathways_matrix = FALSE,
  cores = 1,
  verbose = TRUE
)

Arguments

TrainData

The input training dataset. The first column is the label or the output. For binary classes, 0 and 1 are used to indicate the class member.

TestData

The input test dataset. The first column is the label or the output. For binary classes, 0 and 1 are used to indicate the class member.

pathlistDB

A list of pathways with pathway IDs and their corresponding genes ('entrezID' is used). For details, please refer to ( data("GO2ALLEGS_BP") )

FeatureAnno

The annotation data stored in a data.frame for probe mapping. It must have at least two columns named 'ID' and 'entrezID'. (For details, please refer to data( data("MethylAnno") )

resampling

Resampling in mlr3verse.

nfolds

k-fold cross validation ( Only supported when TestData = NULL )

classifier

Learners in mlr3

predMode

The prediction mode. Available options are c('probability', 'classification').

PathwaySizeUp

The upper-bound of the number of genes in each biological pathways.

PathwaySizeDown

The lower-bound of the number of genes in each biological pathways.

MinfeatureNum_pathways

The minimal defined pathway size after mapping your own data to pathlistDB(KEGG database/GO database).

Add_UnMapped

Whether to add unmapped probes for prediction

Unmapped_num

The number of unmapped probes

Add_FeartureSelection_Method

Feature selection methods.

Inner_CV

Whether to perform a k-fold verification on the training set.

inner_folds

k-fold verification on the training set.

Stage1_FeartureSelection_Method

Feature selection methods.

cutoff

The cutoff used for feature selection threshold. It can be any value between 0 and 1.

Stage2_FeartureSelection_Method

Feature selection methods.

cutoff2

The cutoff used for feature selection threshold. It can be any value between 0 and 1.

classifier2

Learner for stage 2 prediction(if classifier2==NULL,then it is the same as the learner in stage 1.)

target

Is it used to predict or explore potential biological mechanisms? Available options are c('predict', 'pathways').

p.adjust.method

p-value adjustment method.(holm", "hochberg", "hommel", "bonferroni", "BH", "BY",

save_pathways_matrix

Whether to output the path matrix file

cores

The number of cores used for computation.

verbose

Whether to print running process information to the console

Value

A list containing prediction results and prediction result evaluation

Examples

library(mlr3verse)
library(caret)
library(parallel)
library(BioM2)
data=MethylData_Test
set.seed(1)
part=unlist(createDataPartition(data$label,p=0.8))
Train=data[part,]
Test=data[-part,]
pathlistDB=GO2ALLEGS_BP
FeatureAnno=MethylAnno


pred=BioM2(TrainData = Train,TestData = Test,
           pathlistDB=pathlistDB,FeatureAnno=FeatureAnno,
           classifier='svm',nfolds=5,
           PathwaySizeUp=25,PathwaySizeDown=20,MinfeatureNum_pathways=10,
           Add_UnMapped='Yes',Unmapped_num=300,
           Inner_CV='None',inner_folds=5,
           Stage1_FeartureSelection_Method='cor',cutoff=0.3,
           Stage2_FeartureSelection_Method='None',
           target='predict',cores=1
)#(To explore biological mechanisms, set target=‘pathways’)