Skip to contents

Models for predicting clinical and molecular features

Source: R/all.models.R
all.models.Rd

Random forest and glmnet models for predicting various clinical and molecular features of patients diagnosed with prostate cancer. The predictors are gene-level methylation estimated by gene.methylation.

Usage

all.models

Format

An object of class list of length 14.

Details

Models are available for predicting the following features:

  • age.continuous: patient age in years

  • ISUP.grade: International Society of Urological Pathology (ISUP) grade risk group (1-5). See here for further details.

  • t.category: TNM tumour category (1-4), measures the size and extent of the primary tumour

  • psa.categorical: Prostate-specific antigen (PSA) category: <= 10, [10, 20), and >= 20 ng/mL.

  • pga: percentage of the genome altered was defined as PGA = (base-pair length of all genome regions with gain or loss) / 3.2 billion bases * 100

  • <gene>.cna.<loss/gain>: features with .cna. in their name give the gene name and then identify whether there is a copy number loss or gain event. See the Examples section for the full list of cna features.

  • log2p1.snvs.per.mbps: single nucleotide variants (SNVs) per mega-base pairs (Mbps) with a log2(x + 1) transformation.

Examples

data(all.models);

# Models for predicting the following features:
names(all.models);
#>  [1] "age.continuous"       "ISUP.grade"           "t.category"          
#>  [4] "psa.categorical"      "pga"                  "CHD1.cna.loss"       
#>  [7] "NKX3-1.cna.loss"      "MYC.cna.gain"         "PTEN.cna.loss"       
#> [10] "CDKN1B.cna.loss"      "RB1.cna.loss"         "CDH1.cna.loss"       
#> [13] "TP53.cna.loss"        "log2p1.snvs.per.mbps"

# Model class per feature, e.g. randomForest or glmnet:
lapply(all.models, class);
#> $age.continuous
#> [1] "randomForest"
#> 
#> $ISUP.grade
#> [1] "multnet" "glmnet" 
#> 
#> $t.category
#> [1] "randomForest"
#> 
#> $psa.categorical
#> [1] "randomForest"
#> 
#> $pga
#> [1] "randomForest"
#> 
#> $CHD1.cna.loss
#> [1] "lognet" "glmnet"
#> 
#> $`NKX3-1.cna.loss`
#> [1] "randomForest"
#> 
#> $MYC.cna.gain
#> [1] "randomForest"
#> 
#> $PTEN.cna.loss
#> [1] "lognet" "glmnet"
#> 
#> $CDKN1B.cna.loss
#> [1] "lognet" "glmnet"
#> 
#> $RB1.cna.loss
#> [1] "randomForest"
#> 
#> $CDH1.cna.loss
#> [1] "lognet" "glmnet"
#> 
#> $TP53.cna.loss
#> [1] "lognet" "glmnet"
#> 
#> $log2p1.snvs.per.mbps
#> [1] "randomForest"
#> 

# Required genes for predicting each feature:
# lapply(all.models, function(x) x$xNames)