PAD: PAD

In huangwb8/GSClassifier: Gene Signature Classifier

PAD

Description

Unsupervised pan-immune activation/dysfunction (PAD) subtypes of gastric cancer (or other solid tumor) sample based on RNA-Seq/microarray data

Usage

PAD(
  expr,
  PIAM = NULL,
  PIDG = NULL,
  cluster.method = c("ward.D2", "complete", "randomForest")[1],
  rF.para = list(seed = c(2020, 485, 4, 8), ntree = c(300, 300), k = c(2, 2)),
  extra.annot = NULL,
  plot.title = NULL,
  subtype = "PAD.train_20200110",
  verbose = T
)

Arguments

expr	RNA expression matrix. Samples in col and ENSEMBL genes in row.
PIAM	ID of pan-immune activation genes. IF NULL, use default gene set.
PIDG	ID of pan-immune dysfunction genes. IF NULL, use default gene set.
cluster.method	One of 'ward.D2', 'complete' or other methods in hclust. If 'randomForest' was set, the randomForest function would used to classify the samples.
rF.para	The parameters in randomForest function.
extra.annot	Extra top annotation. The same order as colnames of expr.
plot.title	The title of heatmap report.
subtype	Default subtype methods. Now, only 'PAD' and 'ImmuneSubtype' are available. ATTENTION: If you use self-defined data (ens, geneAnnoation, geneSet or scaller), you MUST set subtype as NULL! All available options please visit GSClassifier_Data
verbose	Whether to show heatmap in the process.

Details

This function is used for unsupervised classification of raw data, which is pivotal for the following supervised machine learning. Empirically, the 'ward.D2' method could be useful and high-speed for simple gene signatrues (like PAD classifier). Random forest is a powerful stragety and may act well in larger dataset or complex gene signatures.

Examples

extra.annot = HeatmapAnnotation(
Dataset = id.dataset,
col = list(
  Dataset = if(T){
    l <- mycolor[1:length(unique(id.dataset))];
    names(l) <- unique(id.dataset);
    l}
),
annotation_name_gp = gpar(fontsize = 13, fontface = "bold"),
show_legend = T
)

res1 <- PAD(
  expr = dm.combat.tumor,
  PIAM = piam,
  PIDG = pidg,
  plot.title = 'PanSTAD',
  cluster.method = 'ward.D2',
  subtype = 'PAD.train_20200110',
  extra.annot = extra.annot,
  verbose = T
)

# randomForest: time-consuming in large cohorts
res2 <- PAD(
  expr = dm.combat.tumor,
  PIAM = piam,
  PIDG = pidg,
  cluster.method = 'randomForest',
  rF.para = list(
    seed = c(2020,485,58,152),
    ntree = c(1000,1000),
    k=c(2,2)
  ),
  subtype = 'PAD.train_20200110',
  extra.annot = extra.annot,
  plot.title = 'PanSTAD',
  verbose = T
)

huangwb8/GSClassifier documentation built on July 12, 2024, 5:10 p.m.