Skip to content

Cohort analysis for the identification of driver genes

Key Learning Outcomes

After completing this practical the trainee should be able to:

  • Run the IntOGen analysis software on cohort mutation data.

  • Have gained experience of the structure of the analysis output files in order to identify potential driver genes.

  • Have gained overview knowledge of different methods for identification of genes important in cancers.

Resources You’ll be Using

Tools Used

IntOGen mutations platform:
https://www.intogen.org/search

Sources of Data

TCGA melanoma somatic SNV data from 338 tumour samples:
https://tcga-data.nci.nih.gov/tcga/

Mutation Annotation Format (MAF) specification:
https://wiki.nci.nih.gov/display/TCGA/Mutation+Annotation+Format+(MAF)+Specification

IntOGen installation instructions: https://bitbucket.org/intogen/intogen-pipeline/overview

Author Information

Primary Author(s):
Ann-Marie Patch, QIMR Berghofer ann-marie.patch@qimrberghofer.edu.au
Erdahl Teber, CMRI eteber@cmri.org.au

Contributor(s):
Scott Wood scott.wood@qimrberghofer.edu.au

Introduction

Cancer driver genes are commonly described as genes that when mutated directly affect the potential of a cell to become cancerous. They are important to a tumour cell as they confer a growth or survival advantage over the normal surrounding cells. The mutations in these driver genes are then clonally selected for as the population of tumour cells increases. We think of the key genes driving tumour initiation (development), progression, metastases, resistance and survival. Driver gene mutations are often described as “early” events because they were key in turning a normally functioning and regulated cell into a dysregulated one. The logical assumption is that these key mutations will be present in all tumour cells in a patient’s sample; although sometime this is not true.

There are two major research goals that underline the need to identify driver genes:

  • By identifying the early changes that take place researchers might be able to find a treatment to stop the root cause of why cells become malignant.

  • By identifying groups of patients with the same genes mutated then we can develop therapies that will work for all of them.

When we sequence tumour samples we tend to use samples that come from fully developed cancers that can carry hundreds to thousands of mutations in genes and many more outside of genes. The accumulation of these passenger mutations in cancer cells can happen because often the repair mechanisms or damage sensing processes are amongst the first pathways to become disrupted accelerating the mutational rate. Mutations that occur in genes after the cell has become cancerous may still affect the growth rate, invasiveness and even the response to chemotherapy but may not be present in all cells of a tumour. These genes may be drivers of chemo-resistance or metastasis and are equally good targets for therapies.

IntOGen-mutations is a platform that aims to identify driver mutations using two methodologies from cancer cohort mutation data: the first identifies mutations that are most likely to have a functional impact combined with identifying genes that are frequently mutated; and the second, genes that harbour clustered mutations. These measures are all indicators of positive selection that occurs in cancer evolution and may help the identification of driver genes.

Analysing cancer cohort data with IntOGen

IntOGen-mutations is available as a web based service that can allow users to run their analysis on the host’s servers or it can be downloaded and run on a local server.

For the purposes of the course we will be using a local version of IntOGen so that we don’t encounter any issues sharing resources.

  • To begin open a terminal and navigate to the directory somatic/intogen.

    cd ~/somatic/intogen

In this directory you will find a Mutation Annotation Format (MAF) file containing a cut down version of the somatic variant calls identified from melanoma samples investigated as part of the TCGA cancer genomics projects. You can see what files are in the directory by typing ls, look inside the file using less TCGA_Melanoma_SMgene.maf and close the file and return to the command line by typing q.

  • Run the IntOGen analysis by typing

    intogen -i TCGA_Melanoma_SMgene.maf -o TCGA_Mela_out

The TCGA melanoma maf used in this practical has been modified from the original to reduce processing time and only contains data for the top 680 mutated genes.

The tool will take around 10 minutes to run and the progress will be indicated by the logging lines printed to the terminal. Once complete the output can be explored.

Whilst the tool is running we can explore the options we have used to run IntOGen.

  • To get a list of IntOGen options open up a new terminal

    intogen --help

This command will list the running options that you can alter as command line inputs or in a configuration file. We are using the default options for this run so we didn’t have to supply a configuration file and we only used -i to set the input and -o to control the mane of the output directory.

  • To look at the default options open up the configuration file by typing

    less ~./intogen/task.conf
less ~./intogen/system.conf

It is important to set the correct genome assembly in the task.conf to match the one that you used as your reference when the variant were called. In our task.conf this should be hg19.

Exploring the output of IntOGen

When you run your data over the web on the remote site there is a browse facility that allows you to explore your data using the web version of the database. Running IntOGen locally provides the same tabular information but in a flat file format.

There should be 14 files generated from a successful run of this version of IntOGen:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
gene.tsv
gene.oncodriveclust
pathway.recurrences
gene.oncodrivefm
sample_gene.impact
gene.recurrences                                
sample_variant+transcript.impact
summary.tsv
transcript.recurrences
TCGA_Melanoma_slimSMgene.smconfig
oncodrivefm-pathways-MA_SCORE.tsv
oncodrivefm-pathways-PPH2_SCORE.tsv  
oncodrivefm-pathways-SIFT_SCORE.tsv  
pathway.oncodrivefm

View these files by using ls as below.

1
2
3
```bash
ls ~/somatic/intogen/TCGA_Mela_out/project/TCGA_Melanoma_slimSMgene/
```

This practical will concentrate on the identification of driver genes so we will look at the main output concerning genes. The gene.tsv is the main gene centric output summary table.

  • Open up the gene.tsv file in LibreOffice by double clicking on the icon on your desktop.

  • Select the file tab and click on open.

  • Navigate to the results directory ~/somatic/intogen/TCGA_Mela_out/project/TCGA_Melanoma_slimSMgene/

  • Double click on gene.tsv.

  • In the pop-up box under the Separator options ensure only the tab box is checked and click OK.

This file contains the overall summary results for the IntOGen pipeline presented by gene and reports Q values (i.e. multiple testing corrected P values) for the mutation frequency and cluster modules.

Significantly mutated genes from the cohort data are identified using both the OncodriveFM and OncodriveClUST modules of IntOGen. The OncodriveFM module detects genes that have accumulated mutations with a high functional impact. It uses annotations from the Ensembl variant effect predictor (VEP, V.70) that includes SIFT and Polyphen2 and precomputed MutationAssessor functional impacts. It calculates a P value per gene from the number of mutations detected across all possible coding bases of a gene with a positive weighting for mutations with a high functional impact. The OncodriveCLUST module detects genes that have more variants than would be expected across the cohort that alter the same region of the gene.

The file is sorted to bring the most significantly altered genes to the top. The key columns that help you identify the significantly mutated genes are the 3rd and 4th (C and D) that indicate which of the modules identified a significant result and the Q-values for the modules that are in 21st and 23rd (U and W)

The top twelve genes have significant Q-values for both modules and include BRAF, NRAS and TP53. The next 35 are significant by only one of the modules.

All of these have small Q-values which means they are all significantly mutated genes in this TCGA Melanoma cohort of 338 patients.

  • Now look at their sample frequency count (column 9 MUTS_CS_SAMPLES) these are the number of samples that contain at least one mutation in the gene.

Question

a) Which significantly mutated gene has mutations in the most samples?

b) Which gene/genes have the lowest Q-value from OncodriveFM and OncodriveCLUST?

c) Why don’t the genes with the lowest Q values also have the highest sample frequency value?

Answer

a) BRAF has 175 out of 327 cases with a mutation.

b) TP53 or PTEN have the lowest OncodriveFM Q-values and NRAS has the lowest Q-value for OncodriveCLUST.

c) The P value calculation takes into account the length of the coding sequence of the gene, the mutation rate of the nucleotides contained within it and for OncodriveFM the functional consequences of those changes. Therefore a small gene with a small number of deleterious mutations may have a lower P value and also Q value than a large gene with a high mutation frequency.

The results for the assessment of clustered mutations in genes carried out by the OncodriveCLUST module of IntOGen are shown as amino acid residue positions of the encoded protein.

The three known oncogenes BRAF, NRAS and IDH1 have very low CLUST_QVALUEs indicating that the mutations in these genes are highly clustered. The CLUST_COORDS column reports that there are 160 samples with mutations between the amino acid positions 594-601 of BRAF; 84 samples with mutations at amino acid position 61 of NRAS; and 15 sample with mutations at amino acid position 132 of IDH1.

Question

Why are the oncogenes more likely to have clustered mutations and the tumour suppressor genes less likely?

Answer

Gain of function mutations are required to activate oncogenes and so only key residues in the protein will result in activation. Tumour suppressors are frequently affected by loss of function mutations and deletions. A truncating mutation or frameshift indel can occur in any exon, except the last one, and have the same deleterious functional result.


The other files in the output support the information in this sheet.

The sample_variant+transcript.impact file includes a summary of all mutations found in each of the genes and protein coding transcripts of those genes for all samples identified that have that mutation. It also reports the variant impact scores from SIFT, PolyPhen2, MutationAssesor, reporting also impact categories of which there are four; high, medium, low and none.

  • Open up the sample_variant+transcript.impact file and explore the data.

Question

Can you find out what the nucleotide change details for the most common BRAF mutation that results in V600E amino acid change in the cohort? Sort the data by GENE, then TRANSCRIPT and then PROTEIN_POS to make this easier. The gene ID for BRAF is ENSG00000157764.

Answer

It is an A>T at position chr7:140453136 identified in 127 samples.

References

Gunes et al. Nat. Methods 2010 : http://www.nature.com/nmeth/journal/v7/n2/pdf/nmeth0210-92.pdf

Gonzalez-Perez et al. Nat. Methods 2013 : http://www.nature.com/nmeth/journal/v10/n11/pdf/nmeth.2642.pdf