output
html_document pdf_document default default

Molecular Subtyping High-grade Gliomas

Module authors: Chante Bethell (@cbethell), Stephanie J. Spielman(@sjspielman), and Jaclyn Taroni (@jaclyn-taroni), Jo Lynne Rokita(@jharenza)

Note: The files in the hgg-subset directory were generated via 02-HGG-molecular-subtyping-subset-files.R using the the files in the version 14 data release. When re-running this module, you may want to regenerate the HGG subset files using the most recent data release.

Molecular subtyping HGG workflow

Usage

To run all of the Rscripts in this module from the command line sequentially, use:

bash run-molecular-subtyping-HGG.sh

When run in this manner, 02-HGG-molecular-subtyping-subset-files.R will generate subset files using whichever files are symlinked in data on your local machine.

run-molecular-subtyping-HGG.sh is designed to be run as if it was called from this module directory even when called from outside of this directory.

Folder content

This folder contains scripts tasked to molecularly subtype High-grade Glioma samples in the PBTA dataset. 00-HGG-select-pathology-dx.R gathers the exact matches for inclusion in the pathology_diagnosis and pathology_free_text_diagnosis which are saved in hgg-subset/hgg_subtyping_path_dx_strings.json, which is used downstream in 02-HGG-molecular-subtyping-subset-files to generate subset files.

01-HGG-molecular-subtyping-defining-lesions.Rmd is a notebook written to look at the high-grade glioma defining lesions (H3-3A K28M, H3-3A G35R/V, H3C2 K28M, H3C3 K28M, H3C14 K28M/I) for all tumor samples except LGAT and EPN in the PBTA dataset. This notebook produces a results table found at results/HGG_defining_lesions.tsv.

02-HGG-molecular-subtyping-subset-files.R is a script written to subset the copy number, gene expression, fusion, mutation, SNV and GISTIC's broad values files to include only samples that meet one of the following criteria: 1) with defining lesions 2) have pathology_diagnosis values that match those in hgg-subset/hgg_subtyping_path_dx_strings.json. This script produces the relevant subset files that can be found in the hgg-subset directory.

03-HGG-molecular-subtyping-cnv.Rmd is a notebook written to prepare the copy number data relevant to HGG molecular subtyping. The CNVkit focal copy number file generated in the focal-cn-file-preparation module is used as this CNVkit was also used to produce the GISTIC broad_values_by_arm.txt file that is also implemented in this module. GISTIC arm values are coded as "loss" when arm values are negative, "gain" when arm values are positive, and "neutral" when arm value = 0. For samples that do not have GISTIC results, arm value = NA is given - this will apply to all WXS and panel data since now GISTIC is only run on WGS samples. (Tumor ploidy is not taken into account.) This notebook produces a CNV results table with cleaned CNVkit and GISTIC data found at results/HGG_cleaned_cnv.tsv.

04-HGG-molecular-subtyping-mutation.Rmd is a notebook written to prepare the consensus mutation data relevant to HGG molecular subtyping. We filtered the subset SNV data to the genes of interest based on the following criteria:

• For genes that were mentioned as defining or coocurring lesions (with the exception of TERT; see below), we included only mutations in coding sequences (CDS) that were not classified as silent mutations. Genes with a mutation that met these criteria are stored as comma-separated values in the relevant_coding_mutations column of the cleaned mutation table for a biospecimen.
• Any TERT mutation was included. The Variant_Classification values for any TERT mutation in a biospecimen are included in the TERT_variant_classification column of the cleaned table.
• The IDH1_mutation column of the cleaned table includes the contents of HGVSp_Short when it contains R132 or R172 or No R132 or R172 when no IDH1 mutation that met that criterion was present.

The cleaned table is found at results/HGG_cleaned_mutation.tsv.

05-HGG-molecular-subtyping-fusion.Rmdis a notebook written to prepare the putative oncogenic fusion data relevant to HGG molecular subtyping. Per issue #249, we filtered the data to the two fusions of interest: FGFR1 fusions, which should be mutually exclusive of H3 K28 mutants, and NTRK fusions, which are co-occurring with H3 G35 mutants. Note: NTRK refers to a family of receptor kinases, so we include the full fusion name to account for various individual NTRK gene symbols. Per [issue #474] (d3b-center/ticket-tracker-OPC#474), more gene fusions ( ROS1, ALK, and MET) were included as there is new 2021 entity within HGGs called "Infant-type hemispheric glioma" (IHG). This HGG is cerebral (hemispheric), arises in early childhood, and is characterized by RTK (receptor tyrosine kinase) alterations, typically fusions, in the NTRK family, which has been included before, or in ROS1, ALK, or MET. There is no mention of specific fusion partners or orientations, so we look at any instances of fusions that include the genes mentioned above. This notebook produces a fusion results table found at results/HGG_cleaned_fusion.tsv.

06-HGG-molecular-subtyping-gene-expression.Rmd is a notebook written to prepare the gene expression data relevant to HGG molecular subtyping. Per issue #249, we filtered the z-scored gene expression to genes of interest: OLIG2 and FOXG1 should be highly expressed in IDH mutants, and TP73-AS1 methylation and downregulation cooccurs with TP53 mutations. This notebook produces two expression results table (one for each selection strategy) found at results/HGG_cleaned_expression.tsv.

07-HGG-molecular-subtyping-combine-table.Rmd is a notebook written to combine the cleaned copy number, mutation, fusion, and gene expression data (prepared in this module's previous notebooks) into one final table of results. This notebook produces one table with the cleaned data found at results/HGG_cleaned_all_table.tsv. Methylation classification is used during subtyping. The DKFZ v12b6 data are available in dkfz_v12_methylation_subclass and subtypes with dkfz_v12_methylation_subclass_score >= 0.8 are considered high-confidence and used here. The NIH Bethesda classifier v2 data are available in NIH_v2_methylation_Class and subtypes with NIH_v2_methylation_Class_mean_score >= 0.9 and NIH_v2_methylation_Superfamily_mean_score >= 0.9 are considered high-confidence and determined in order using DKFZ first and if a subtype score is < 0.8 for DKFZ but is high-confidence for NIH, then the NIH subtype is used.

A table with the molecular subtype information for each HGG sample at results/HGG_molecular_subtype.tsv is also produced, where the subtype values in the molecular_subtype column are determined as follows:

1. If there was an H3-3A K28M or K28I, H3C2 K28M or K28I, H3C3 K28M or K28I, or H3C14 K28M or K28I mutation or high-confidence methylation subtype (DMG_K27, DMG_EGFR, or GBM_THAL(K27))-> DMG, H3K28
2. If there was an H3-3A G35V or G35R mutation or high-confidence methylation subtype (DHG_G34 or GBM_G34) -> HGG, H3 G35
3. If there was an IDH1 R132 mutation or high-confidence methylation subtype (A_IDH_HG or GBM_IDH)-> HGG, IDH
4. In histologies_base.tsv, column pathology_free_text_diagnosis contains "infant type hemispheric glioma" or dkfz_v12_methylation_subclass == IHG -> IHG
   1. If there was a NTRK fusion -> IHG, NTRK-altered
   2. If there was a ROS1 fusion -> IHG, ROS1-altered
   3. If there was a ALK fusion -> IHG, ALK-altered
   4. If there was a MET fusion -> IHG, MET-altered
   5. If there was no fusion -> IHG, To be classified based on IHG methylation classification and sample clinical report in the pathology_diagnosis_free_text stated as infant type hemispheric glioma
5. If methylation subtype == "PXA" or pathology_free_text_diagnosis contains "pleomorphic xanthoastrocytoma" or "pxa" AND there was a BRAF V600E mutation AND loss of CDKN2A and/or CDKN2B -> HGG, PXA
6. If methylation subtype == "O_IDH" -> Oligo, IDH
7. If methylation subtype == "OLIGO_IDH" -> Oligosarcoma, IDH
8. All other samples that did not meet any of these criteria were marked as HGG, H3 wildtype if there was no canonical histone variant the DNA sample, the methylation classification subtype if present, or else HGG, To be classified

08-1p19q-codeleted-oligodendrogliomas.Rmd is a notebook written to identify samples in the OpenPBTA dataset that should be classified as 1p/19q co-deleted oligodendrogliomas. The GISTIC broad_values_by_arm.txt file is used to identify samples with 1p and 19q loss, then the consensus mutation file is filtered to the identified samples in order to check for IDH1 mutations. Note: Per this comment, very few samples in the OpenPBTA dataset, if any, are expected to fit into the 1p/19q co-deleted oligodendrogliomas subtype. Also please NOTE: we currently only have GISTIC scores for WGS so we cannot tell whether a sample is 1p/19q co-deleted oligodendrogliomas if they are sequenced with WXS or targeted sequencing.

09-HGG-with-braf-clustering.Rmd is a notebook written to identify high grade glioma samples without histone mutations that have BRAF V600E mutations and observe how they cluster alongside low grade gliomas and high grade gliomas without the BRAF V600E mutation in the stranded RNA-seq data (which contains both histologies) in RNA-seq data. We plotted the t-SNE and UMAP results from the transcriptomic-dimension-reduction, highlighting samples without histone mutations and with a BRAF V600E mutation. The results, shown below, suggest that one sample may be a candidate for reclassification (also saved in the plots directory of this module).

Folder structure

The structure of this folder is as follows:

├── 00-HGG-select-pathology-dx.Rmd
├── 00-HGG-select-pathology-dx.nb.html
├── 01-HGG-molecular-subtyping-defining-lesions.Rmd
├── 01-HGG-molecular-subtyping-defining-lesions.nb.html
├── 02-HGG-molecular-subtyping-subset-files.R
├── 03-HGG-molecular-subtyping-cnv.Rmd
├── 03-HGG-molecular-subtyping-cnv.nb.html
├── 04-HGG-molecular-subtyping-mutation.Rmd
├── 04-HGG-molecular-subtyping-mutation.nb.html
├── 05-HGG-molecular-subtyping-fusion.Rmd
├── 05-HGG-molecular-subtyping-fusion.nb.html
├── 06-HGG-molecular-subtyping-gene-expression.Rmd
├── 06-HGG-molecular-subtyping-gene-expression.nb.html
├── 07-HGG-molecular-subtyping-combine-table.Rmd
├── 07-HGG-molecular-subtyping-combine-table.nb.html
├── 08-1p19q-codeleted-oligodendrogliomas.Rmd
├── 08-1p19q-codeleted-oligodendrogliomas.nb.html
├── 09-HGG-with-braf-clustering.Rmd
├── 09-HGG-with-braf-clustering.nb.html
├── README.md
├── hgg-subset
│   ├── hgg_focal_cn.tsv.gz
│   ├── hgg_fusion.tsv
│   ├── hgg_gistic_broad_values.tsv
│   ├── hgg_metadata.tsv
│   ├── hgg_snv_maf.tsv.gz
│   ├── hgg_subtyping_path_dx_strings.json
│   └── hgg_zscored_expression.RDS
├── plots
│   ├── HGG_stranded.pdf
│   ├── HGG_stranded.png
│   └── mol_subtype_workflow.png
├── results
│   ├── HGG_cleaned_all_table.tsv
│   ├── HGG_cleaned_cnv.tsv
│   ├── HGG_cleaned_expression.tsv
│   ├── HGG_cleaned_fusion.tsv
│   ├── HGG_cleaned_mutation.tsv
│   ├── HGG_defining_lesions.tsv
│   └── HGG_molecular_subtype.tsv
└── run-molecular-subtyping-HGG.sh