DNA methylation-based classification of central nervous system tumours

Collection of R-scripts used to perform DNA-methylation data analysis presented in DNA methylation-based classification of central nervous system tumours. The raw data is publicly available at NCBI GEO under Accession number GSE90496.

Preprocessing and Normalization

preprocessing.R

Reads raw data, performs normalization, basic filtering and batch effect adjustment between Frozen and FFPE samples. Normalized and batch adjusted as well as unadjusted data is stored in ./results

Unsupervised tSNE analysis

tsne.R

Performs non-linear dimension reduction on preprocessed DNA-methylation data.

Classifier training and cross-validation

training.R

Trains the Random Forest classifier on the complete data set and stores the final classifier in ./results

cross_validation.R

Performs nested cross-validation and stores the results in ./CV

calibration.R

Evaluates the results of the cross-validation and fits a calibration model that is stored in ./results and compiles a final report showing classifier performance metrics CVresults.html.

Tumor purity estimation

purity.Rmd
purity.html

Example how TCGA 450k methylation data and ABSOLUTE tumor purity estimates can be used to train a Random Forest to predict tumor purity.

Copy number analysis

cnv_analysis.R

Example how the conumee Bioconductor is applied to perform copy number variation analysis. Note, to get the reference objects stored in ./CNV_data, Git large file storage needs to be installed before cloning the repository.