Identification of somatic and germline variants from tumor and normal sample pairs

Overview

question Questions
  • What are the specific challenges in somatic variant calling that set it apart from regular diploid variant calling?
  • How can you call variants and classify them according to their presence/absence in/from tumor and normal tissue of the same individual?
  • How can you annotate variants and affected genes with prior knowledge from human genetic and cancer-specific databases to generate clinically relevant reports?
objectives Objectives
  • Call variants and their somatic status from whole-exome sequencing data
  • Annotate variants with a wealth of human genetic and cancer-specific information extracted from public databases
  • Add gene-level annotations and generate reports of annotated somatic and germline variants, loss-of-heterozygosity (LOH) events, and affected genes, ready for interpretation by clinicians
requirements Requirements

time Time estimation: 7 hours

Supporting Materials

Introduction

When sequencing genomic material from a human tumor, the underlying clinical or research question typically is what spectrum of mutations distinguishes this tumor from healthy tissue.

This question cannot be answered adequately just by comparing the tumor tissue to the human reference genome because even in healthy tissue there will be many thousands of variants compared to the reference genome. This is because every individual inherits a unique pattern of that many variants from her parents. A fundamental difference between these variants and the tumor-specific mutations is that the former are present in the carrier’s germline, while the latter have been acquired somatically and will, thus, not be transmitted to offspring. Therefor, we talk of germline variants to refer to variants present in healthy and tumor tissue alike, and of somatic variants to refer to tumor-specific variants. To be able to distinguish between these two types of variants always requires a direct comparison of data from tumor and normal tissue samples.

In addition to acquiring new variants, tumors can also lose or gain chromosomal copies of variants found heterozygously in an individual’s germline. This phenomenon is termed loss of heterozygosity (LOH) because only one of the two original alleles persists in the tumor (either in a hemizygous state if the other allele is simply dropped, or in a homozygous state in the case of a duplication of one allelic copy accompanied by loss of the other). The detection of LOH events, again, is dependent on a comparison of tumor and normal tissue data.

In this tutorial we are going to identify somatic and germline variants, as well as variants affected by LOH, from a tumor and a normal sample of the same patient. Our goal is to report the variant sites, and the genes affected by them, annotated with the content of general human genetic and cancer-specific databases. Ideally, this may provide insight into the genetic events driving tumor formation and growth in the patient, and might be of prognostic and even therapeutic value by revealing variants known to affect drug resistance/sensitivity, tumor aggressiveness, etc.

Agenda

In this tutorial, we will cover:

  1. Data Preparation
    1. Get data
  2. Quality control and mapping of NGS reads
    1. Quality control
    2. Read trimming and filtering
    3. Read Mapping
  3. Mapped reads postprocessing
    1. Filtering on mapped reads properties
    2. Removing duplicate reads
    3. Left-align reads around indels
    4. Recalibrate read mapping qualities
  4. Variant calling and classification
  5. Variant annotation and reporting
    1. Get data
    2. Adding annotations to the called variants
    3. Reporting selected subsets of variants
    4. Generating reports of genes affected by variants
    5. Adding additional annotations to the gene-centered report

Data Preparation

First we need to upload and prepare some input data to analyze. The sequencing reads we are going to analyze are from real-world data from a cancer patient’s tumor and normal tissue samples. For the sake of an acceptable speed of the analysis, the original data has been downsampled though to include only the reads from human chromosomes 5, 12 and 17.

Get data

hands_on Hands-on: Data upload

  1. Create a new history for this tutorial and give it a meaningful name

    tip Tip: Creating a new history

    Click the new-history icon at the top of the history panel

    If the new-history is missing:

    1. Click on the galaxy-gear icon (History options) on the top of the history panel
    2. Select the option Create New from the menu

    tip Tip: Renaming a history

    1. Click on Unnamed history (or the current name of the history) (Click to rename history) at the top of your history panel
    2. Type the new name
    3. Press Enter
  2. Import the following four files from Zenodo:

    https://zenodo.org/record/2582555/files/SLGFSK-N_231335_r1_chr5_12_17.fastq.gz
    https://zenodo.org/record/2582555/files/SLGFSK-N_231335_r2_chr5_12_17.fastq.gz
    https://zenodo.org/record/2582555/files/SLGFSK-T_231336_r1_chr5_12_17.fastq.gz
    https://zenodo.org/record/2582555/files/SLGFSK-T_231336_r2_chr5_12_17.fastq.gz
    

    where the first two files represent the forward and reverse reads sequence data from a patient’s normal tissue, and the last two represent the data of the same patient’s tumor tissue.

    Alternatively, the same files may be available on your Galaxy server through a shared data library (your instructor may tell you so), in which case you may prefer to import the data directly from there.

    • Copy the link location
    • Open the Galaxy Upload Manager (galaxy-upload on the top-right of the tool panel)

    • Select Paste/Fetch Data
    • Paste the link into the text field

    • Change Type from “Auto-detect” to fastqsanger.gz

    • Press Start

    • Close the window

    By default, Galaxy uses the URL as the name, so rename the files with a more useful name.

    tip Tip: Importing data from a data library

    As an alternative to uploading the data from a URL or your computer, the files may also have been made available from a shared data library:

    • Go into Shared data (top panel) then Data libraries

    • Find the correct folder (ask your instructor)

    • Select the desired files
    • Click on the To History button near the top and select as Datasets from the dropdown menu
    • In the pop-up window, select the history you want to import the files to (or create a new one)
    • Click on Import
  3. Check that all newly created datasets in your history have their datatypes assigned correctly, and fix any missing or wrong datatype assignment

    tip Tip: Changing the datatype

    • Click on the galaxy-pencil pencil icon for the dataset to edit its attributes
    • In the central panel, click on the galaxy-chart-select-data Datatypes tab on the top
    • Select fastqsanger.gz
    • Click the Change datatype button
  4. Rename the datasets and add appropriate tags to them

    For datasets that you upload via a link, Galaxy will pick the link address as the dataset name, which you will likely want to shorten to just the file names.

    tip Tip: Renaming a dataset

    • Click on the galaxy-pencil pencil icon for the dataset to edit its attributes
    • In the central panel, change the Name field
    • Click the Save button

    Large parts of the analysis in this tutorial will consist of identical steps performed on the normal and on the tumor tissue data in parallel.

    To make it easier to keep track of which dataset represents which branch of an analysis in a linear history, Galaxy supports dataset tags. In particular, if you attach a tag starting with # to any dataset, that tag will automatically propagate to any new dataset derived from the tagged dataset.

    tip Tip: Adding a tag

    • Click on the dataset
    • Click on galaxy-tags Edit dataset tags
    • Add a tag starting with #

      Tags starting with # will be automatically propagated to the outputs of tools using this dataset.

    • Check that the tag is appearing below the dataset name

    Before starting our analysis it is, thus, a good idea to tag the two fastq datasets representing the normal tissue with, e.g., #normal and the two datasets representing the tumor tissue with, e.g., #tumor.

  5. Import the reference genome with the hg19 version of the sequences of human chromosomes 5, 12 and 17:

    https://zenodo.org/record/2582555/files/hg19.chr5_12_17.fa.gz
    

    Make sure you specify the datatype as fasta in the import dialog.

    comment Shortcut

    You can skip this step if the Galaxy server you are working on offers a hg19 version of the human reference genome with prebuilt indexes for bwa-mem and samtools (ask your instructor, or check the tools Map with BWA-MEM tool and VarScan Somatic tool if they list a hg19 version as an option under “(Using) reference genome”).

    Alternatively, load the dataset from a shared data library.

  6. Rename the reference genome

    The reference genome you have imported above came as a compressed file, but got unpacked by Galaxy to plain fasta format according to your datatype selection. You may now wish to remove the .gz suffix from the dataset name.

Quality control and mapping of NGS reads

Before starting our analysis, we would like to make sure that the input data is of good quality, i.e., that there haven’t been any major issues during DNA preparation, exon capture, or during actual sequencing. To avoid spurious variant calls due to low input quality, we can ensure that all sequencing reads used in the analysis meet some minimal quality criteria by trimming low-quality parts off of the ends of reads and/or discarding reads of poor quality altogether. The resulting set of polished reads then needs to be mapped to the human reference genome because knowing the genomic positions that the bases of a read provide evidence for is, of course, a prerequisite for variant calling.

comment More on quality control and mapping

If you would like to explore the topics of quality control and read mapping in detail, you should take a look at the separate Quality Control and Mapping tutorials. Here, we will only illustrate the concrete steps necessary for quality control and read mapping of our particular datasets.

Quality control

hands_on Hands-on: Quality control of the input datasets

  1. Run FastQC tool on each of your four fastq datasets
    • param-files “Short read data from your current history”: all 4 FASTQ datasets selected with Multiple datasets

    tip Tip: Select multiple datasets

    1. Click on param-files Multiple datasets
    2. Select several files by keeping the Ctrl (or COMMAND) key pressed and clicking on the files of interest

    When you start this job, eight new datasets (one with the calculated raw data, another one with an html report of the findings for each input dataset) will get added to your history.

  2. Use MultiQC tool to aggregate the raw FastQC data of all four input datasets into one report
    • In “Results”
      • “Which tool was used generate logs?”: FastQC
      • In “FastQC output”
        • “Type of FastQC output?”: Raw data
        • param-files “FastQC output”: all four RawData outputs of FastQC tool)
  3. Inspect the Webpage output produced by the tool

    question Questions

    1. What do you think of the base qualities of the sequences?
    2. Which aspect of the quality report is most puzzling to you? (Hint: Have a look at the GC content of the reads)

    solution Solution

    1. The quality of the sequences looks promising. There are no discernible systematic problems with it.

      Even the reverse reads, which are typically of somewhat poorer quality than the corresponding forward reads, look good on average.

    2. The GC content plots of the forward and the reverse reads from both samples reveal a very peculiar bimodal distribution.

      Typically, a non-normal distribution of the GC content of the reads from a sample is considered to hint at possible contamination. Here, however, we are dealing with sequencing data from captured exomes, i.e, the reads are not representing random sequences from a genome, but rather an arbitrary selection. In fact, the samples at hand were prepared using Agilent’s SureSelect V5 technology for exome enrichment, and bimodal GC content distributions can be seen as a hallmark of that capture method in several publications (see, for example, Fig. 4C in Meienberg et al., 2015 ).

Read trimming and filtering

Although the raw reads used in this tutorial are of relatively good overall quality already, we will apply read trimming and filtering to see if we can improve things still a bit, but also to demonstrate the general concept.

hands_on Hands-on: Read trimming and filtering of the normal tissue reads

  1. Trimmomatic tool to trim and filter the normal tissue reads
    • “Single-end or paired-end reads?”: Paired-end (two separate input files)

      This makes the tool treat the forward and reverse reads simultaneously.

      • param-file “Input FASTQ file (R1/first of pair)”: the forward reads (r1) dataset of the normal tissue sample
      • param-file “Input FASTQ file (R2/second of pair)”: the reverse reads (r2) dataset of the normal tissue sample
    • “Perform initial ILLUMINACLIP step?”: Yes
      • “Select standard adapter sequences or provide custom?”: Standard
        • “Adapter sequences to use”: TruSeq3 (paired-ended, for MiSeq and HiSeq)
      • “Maximum mismatch count which will still allow a full match to be performed”: 2
      • “How accurate the match between the two ‘adapter ligated’ reads must be for PE palindrome read alignment”: 30
      • “How accurate the match between any adapter etc. sequence must be against a read”: 10
      • “Minimum length of adapter that needs to be detected (PE specific/ palindrome mode)”: 8
      • “Always keep both reads (PE specific/palindrome mode)?”: Yes

      These parameters are used to cut ILLUMINA-specific adapter sequences from the reads.

    • In “Trimmomatic Operation”
      • In “1: Trimmomatic Operation”
        • “Select Trimmomatic operation to perform”: Cut the specified number of bases from the start of the read (HEADCROP)
          • “Number of bases to remove from the start of the read”: 3
      • param-repeat “Insert Trimmomatic Operation”*
      • In “2: Trimmomatic Operation”
        • “Select Trimmomatic operation to perform”: Cut bases off the end of a read, if below a threshold quality (TRAILING)
          • “Minimum quality required to keep a base”: 10
      • param-repeat “Insert Trimmomatic Operation”*
      • In “3: Trimmomatic Operation”
        • “Select Trimmomatic operation to perform”: Drop reads below a specified length (MINLEN)
          • “Minimum quality required to keep a base”: 25

      These three trimming and filtering operations will be applied to the reads in the given order after adapter trimming.

Running this job will generate four output datasets:

  • two for the trimmed forward and reverse reads that still have a proper mate in the other dataset
  • two more datasets of orphaned forward and reverse reads, for which the corresponding mate got dropped because of insufficient length after trimming; when you inspect these two files, however, you should find that they are empty because none of our relatively high quality reads got trimmed that excessively. You can delete the two datasets to keep your history more compact.

details More on handling unpaired reads in paired-end data

Splitting out potential unpaired reads into separate datasets like this is important because read mappers, typically, expect reads in forward and reverse input datasets to be arranged in proper pairs, and reads in one of the datasets without a counterpart in the other would destroy that expected structure. Therefor, when your data is paired-end data always make sure you use Trimmomatic in paired-end mode, too!

hands_on Hands-on: Read trimming and filtering of the tumor tissue reads

  1. Trim and filter the tumor tissue reads following the same steps as above, just change the two input datasets to treat the tumor tissue reads with identical settings.

  2. Check that the two unpaired reads datasets are empty, and delete them.

hands_on Exercise: Quality control of the polished datasets

Use FastQC tool and MultiQC tool like before, but using the four trimmed datasets produced by Trimmomatic as input.

question Questions

How did read trimming affect the quality reports?

solution Solution

As expected, trimming the relatively high-quality raw reads did not have any substantial impact on average. A small fraction of successful adapter removal is visible though.

Read Mapping

hands_on Hands-on: Read Mapping

  1. Map with BWA-MEM tool to map the reads from the normal tissue sample to the reference genome
    • “Will you select a reference genome from your history or use a built-in index?”: Use a built-in genome index
      • “Using reference genome”: Human: hg19 (or a similarly named option)

      comment Using the imported hg19 sequence

      If you have imported the hg19 sequence as a fasta dataset into your history instead:

      • “Will you select a reference genome from your history or use a built-in index?”: Use a genome from history and build index
        • param-file “Use the following dataset as the reference sequence”: your imported hg19 fasta dataset.
    • “Single or Paired-end reads”: Paired
      • param-file “Select first set of reads”: the trimmed forward reads (r1) dataset of the normal tissue sample; output of Trimmomatic tool
      • param-file “Select second set of reads”: the trimmed reverse reads (r2) dataset of the normal tissue sample; output of Trimmomatic tool
    • “Set read groups information?”: Set read groups (SAM/BAM specification)
      • “Auto-assign”: No
        • “Read group identifier (ID)”: 231335 (this value being taken from the original name of the normal tissue input files)
      • “Auto-assign”: No
        • “Read group sample name (SM)”: Normal

    comment More on read group identifiers and sample names

    In general, you are free to choose ID and SM values to your liking, but the ID should unambiguously identify the sequencing run that produced the reads, while the SM value should identify the biological sample.

  2. Map with BWA-MEM tool to map the reads from the tumor tissue sample,
    • “Will you select a reference genome from your history or use a built-in index?”: Use a built-in genome index
      • “Using reference genome”: Human: hg19 (or a similarly named option)

      Adjust these settings as before if you are using the imported reference genome.

    • “Single or Paired-end reads”: Paired
      • param-file “Select first set of reads”: the trimmed forward reads (r1) dataset of the tumor tissue sample; output of Trimmomatic tool
      • param-file “Select second set of reads”: the reverse reads (r2) dataset of the tumor tissue sample; output of Trimmomatic tool
    • “Set read groups information?”: Set read groups (SAM/BAM specification)
      • “Auto-assign”: No
        • “Read group identifier (ID)”: 231336 (this value, again, being taken from the original name of the tumor tissue input files)
      • “Auto-assign”: No
        • “Read group sample name (SM)”: Tumor

Mapped reads postprocessing

To ensure that we base our variant analysis only on unambiguous, high-quality read mappings we will do some postprocessing next.

Filtering on mapped reads properties

To produce new filtered BAM datasets with only those reads retained that have a minimal mapping quality of 1 and are mapped in a proper pair:

hands_on Hands-on: Filtering for mapping status and quality

  1. Filter SAM or BAM, output SAM or BAM tool with the following parameters (leaving non-mentioned ones at their defaults):
    • param-files “SAM or BAM file to filter”: mapped reads datasets from the normal and the tumor tissue data, outputs of Map with BWA-MEM tool
    • “Minimum MAPQ quality score”: 1
    • “Filter on bitwise flag”: yes
    • “Only output alignments with all of these flag bits set”:
      • param-check “Read is mapped in a proper pair”

This will result in two new datasets, one for each of the normal and tumor data.

Removing duplicate reads

hands_on Hands-on: Remove duplicates

  1. RmDup tool with the following parameters:
    • param-files “BAM file”: filtered reads datasets from the normal and the tumor tissue data; the outputs of Filter SAM or BAM
    • “Is this paired-end or single end data”: BAM is paired-end
    • “Treat as single-end”: No

Again, this will produce two new datasets, one for each of the normal and tumor data.

Left-align reads around indels

hands_on Hands-on: Left-align

  1. BamLeftAlign tool with the following parameters:
    • “Choose the source for the reference genome”: Locally cached
      • param-files “BAM dataset to re-align”: your filtered and deduplicated reads datasets from the normal and the tumor tissue data; the outputs of RmDup
      • “Using reference genome”: Human: hg19 (or a similarly named choice) - “Maximum number of iterations”: 5

comment Using the imported hg19 sequence

If you have imported the hg19 sequence as a fasta dataset into your history instead:

  • “Choose the source for the reference genome”: History
    • param-file “Using reference file”: your imported hg19 fasta dataset

As before, this will generate two new datasets, one for each of the normal and tumor data.

Recalibrate read mapping qualities

hands_on Hands-on: Recalibrate read quality scores

  1. CalMD tool with the following parameters:
    • param-files “BAM file to recalculate”: the left-aligned datasets from the normal and the tumor tissue data; the outputs of BamLeftAlign tool
    • “Choose the source for the reference genome”: Use a built-in genome
      • “Using reference genome”: Human: hg19 (or a similarly named choice)

      comment Using the imported hg19 sequence

      If you have imported the hg19 sequence as a fasta dataset into your history instead:

      • “Choose the source for the reference genome”: Use a genome from the history
        • param-file “Using reference file”: your imported hg19 fasta dataset.
    • “Do you also want BAQ (Base Alignment Quality) scores to be calculated?”: No

      The VarScan somatic tool that we are going to use for calling variants at the next step is typically used in combination with unadjusted base quality scores because the general opinion is that the base quality downgrades performed by CalMD and other tools from the samtools suite of tools are too severe for VarScan to retain good sensitivity. We are sticking with this practice in this tutorial.

      comment Using adjusted base quality scores

      If, for your own data, you would like to experiment with adjusted base quality scores, it is important to understand that VarScan somatic will only make use of the adjusted scores if they are incorporated directly into the read base qualities of a BAM input dataset, but not if they are written to the dataset separately.

      Hence, should you ever decide to use:

      • “Do you also want BAQ (Base Alignment Quality) scores to be calculated?”: Yes, run BAQ calculation

        and you want this setting to affect downstream variant calling with VarScan somatic, make sure you also set then:

        • “Use BAQ to cap read base qualities”: Yes

      Please also note that BAQ scores are quite expensive to calculate so be prepared to see a substantial (up to 10x!) increase in job run time when enabling it.

    • “Additional options”: Advanced options
      • “Change identical bases to ‘=’“: No
      • “Coefficient to cap mapping quality of poorly mapped reads”: 50

        This last setting is the real reason why we use CalMD at this point. It is an empirical, but well-established finding that the mapping quality of reads mapped with bwa should be capped this way before variant calling.

This will, once more, produce two new datasets, one for each of the normal and tumor data.

Variant calling and classification

Having generated a high-quality set of mapped read pairs, we can proceed to variant calling. The tool VarScan somatic is a dedicated solution for somatic variant calling that:

  • detects variant alleles in tumor/normal sample pairs
  • calls sample genotypes at variant sites
  • classifies variants into germline, somatic and LOH event variants using solid classical statistics even in the presence of non-pure samples like those obtained from biopsies

hands_on Hands-on: Variant calling and classification

  1. VarScan somatic tool with the following parameters:
    • “Will you select a reference genome from your history or use a built-in genome?”: Use a built-in genome
      • “reference genome”: Human: hg19 (or a similarly named choice)

      comment Using the imported hg19 sequence

      If you have imported the hg19 sequence as a fasta dataset into your history instead:

      • “Will you select a reference genome from your history or use a built-in genome?”: Use a genome from the history
        • param-file “reference genome”: your imported hg19 fasta dataset.
    • param-file “aligned reads from normal sample”: the mapped and fully post-processed normal tissue dataset; one of the two outputs of CalMD tool
    • param-file“aligned reads from tumor sample”: the mapped and fully post-processed tumor tissue dataset; the other output of CalMD tool
    • “Estimated purity (non-tumor content) of normal sample”: 1
    • “Estimated purity (tumor content) of tumor sample”: 0.5
    • “Generate separate output datasets for SNP and indel calls?”: No
    • “Settings for Variant Calling”: Customize settings
      • “Minimum base quality”: 28

        We have seen, at the quality control step, that our sequencing data is of really good quality, and we have chosen not to downgrade base qualities at the quality scores recalibration step above, so we can increase the base quality required at any given position without throwing away too much of our data.

      • “Minimum mapping quality”: 1

        During postprocessing, we have filtered our reads for ones with a mapping quality of at least one, so requiring this quality also here does not actually change the results, but it makes the requirement more explicit.

      Leave all other settings in this section at their default values.

    • “Settings for Posterior Variant Filtering”: Use default values

Variant annotation and reporting

For this tutorial we are going to use variant and gene annotations from many different sources. Most of these are handled for us by the tools we are going to use in this section, but we need to take care of importing the data from four sources into Galaxy separately:

  • variant annotations from Cancer Hotspots
  • variant and gene information from the Cancer Biomarkers database of the Cancer Genome Interpreter (CGI) project
  • variant and gene information from the CIViC database
  • variant annotations from dbSNP
  • lists of genes annotated with the keywords proto-oncogene or tumor suppressor at UniProt

Each of these annotation sets has been released either in the public domain or under a free data license, which allows you to use them as part of this tutorial, but also for other purposes.

Starting from the data downloaded from each of these sites, we have generated a set of new data files tailored to the requirements of the workflow of this tutorial and have made them available through Zenodo, again under a free data license.

Get data

hands_on Hands-on: Data upload

  1. Import the following variant annotation files from Zenodo:

    https://zenodo.org/record/2581873/files/hotspots.bed
    https://zenodo.org/record/2581873/files/cgi_variant_positions.bed
    https://zenodo.org/record/2581873/files/01-Feb-2019-CIVic.bed
    https://zenodo.org/record/2582555/files/dbsnp.b147.chr5_12_17.vcf.gz
    

    Make sure you select bed as the datatype for the first three files and vcf for the last file.

    Alternatively, add the files from a shared data library on your Galaxy server instance.

  2. Import some gene-level annotation files from Zenodo:

    https://zenodo.org/record/2581881/files/Uniprot_Cancer_Genes.13Feb2019.txt
    https://zenodo.org/record/2581881/files/cgi_genes.txt
    https://zenodo.org/record/2581881/files/01-Feb-2019-GeneSummaries.tsv
    

    and make sure you select tabular as their datatype, or add them from the shared data library.

  3. Download SnpEff functional genomic annotations

    comment Shortcut

    You can skip this step if the Galaxy server you are working on offers Homo sapiens: hg19 as a locally installed snpEff database. You can check the SnpEff eff tool tool under Genome source to see if this is the case.

    Use SnpEff Download tool to download genome annotation database hg19.

  4. Check that the datatypes of all new datasets have been set correctly, and change them if necessary. You may also want to shorten some of the dataset names.

Adding annotations to the called variants

Adding functional genomic annotations

Certainly, not all variants are equal. Many may just be silent mutations with no effect at the amino acid level, while a few others may be disrupting the coding sequence of a protein by introducing a premature stop codon or a frameshift. Of course, it is also important to know which gene is affected by a variant. Such functional genomic annotations can be added to a VCF dataset of variants with SnpEff.

hands_on Hands-on: Adding annotations with SnpEff

  1. SnpEff eff tool
    • param-file “Sequence changes (SNPs, MNPs, InDels)”: the output of VarScan somatic tool
    • “Input format”: VCF
    • “Output format”: VCF (only if input is VCF)
    • “Genome source”: Locally installed reference genome
      • “Genome”: Homo sapiens: hg19 (or a similarly named option)

      comment Using the imported hg19 SnpEff genome database

      If you have imported the hg19 SnpEff genome database into your history instead:

      • “Genome source”: Downloaded snpEff database in your history
        • param-file “SnpEff4.3 Genome Data”: your imported hg19 SnpEff dataset.
    • “Produce Summary Stats”: No

Adding genetic and clinical evidence-based annotations

Other interesting pieces of information about a variant include aspects like whether the variant has been observed before in the human population and, if so, at which frequency. If a variant is known to be associated with specific diseases, we would also very much like to know that. To proceed with this kind of genetic and clinical evidence-based annotations, we are going to convert our list of variants into a database that can be handled more efficiently than a VCF dataset. We will use the GEMINI tool suite for this task and for all further work with the variants.

hands_on Hands-on: Creating a GEMINI database from a variants dataset

  1. GEMINI load tool with
    • param-file “VCF dataset to be loaded in the GEMINI database”: the output of SnpEff eff tool
    • “The variants in this input are”: annotated with snpEff
    • “This input comes with genotype calls for its samples”: Yes

      Calling sample genotypes was part of what we used VarScan somatic for.

    • “Choose a gemini annotation source”: select the latest available annotations snapshot (most likely, there will be only one)
    • “Sample and family information in PED format”: Nothing selected
    • “Load the following optional content into the database”
      • param-check “GERP scores”
      • param-check “CADD scores”
      • param-check “Gene tables”
      • param-check “Sample genotypes”
      • param-check “variant INFO field”

      Be careful to leave unchecked:

      • “Genotype likelihoods (sample PLs)”

        VarScan somatic does not generate these values

      • “only variants that passed all filters”

        It is simple and more flexible to filter for this property later

During the creation of the database GEMINI already (silently) adds an impressive amount of annotations it knows about to our variants (including, e.g., the frequency at which every variant has been observed in large human genome sequencing projects). We have gotten all of these for free, just by converting the variants to a GEMINI database!

GEMINI also extracts a lot of the information stored in the VCF input dataset for us (such as the functional genomic annotations that we added with SnpEff).

However, there are typically additional annotations from other sources (not incorporated into GEMINI) that one would like to add. In addition, GEMINI is not prepared to extract some non-standard information from VCF datasets, including some important bits added by VarScan somatic.

GEMINI annotate tool is the tool that is designed to help you with these rather common issues. It lets you add further annotations to the variants in an already loaded GEMINI database.

As a first step, we are going to use the tool to add some crucial information generated by VarScan somatic, but not recognized by GEMINI load, to the database. Specifically, we are interested in three values calculated by VarScan somatic for each variant it detected:

  • Somatic status (SS)

    This is a simple numeric code, in which a value of 1 indicates a germline variant, 2 a somatic variant and 3 an LOH event.

  • Germline p-value (GPV)

    For variants with a somatic status of 1, this is the error probability associated with that status call.

  • Somatic p-value (SPV)

    This is the error probability associated with status calls of 2 and 3 (somatic and LOH calls).

These values are encoded in the INFO column of the VarScan-generated VCF dataset and we are going to extract them from there and add them to the GEMINI database.

details Handling of the VCF INFO field by GEMINI load

If you paid close attention to how we generated the GEMINI database, you might remember that, under “Load the following optional content into the database”, we checked the option param-check “variant INFO field”. Did the tool not do what it was supposed to do?

The answer is that it did, but not in the way you may expect.

In fact, GEMINI load always extracts INFO field elements it knows about and stores them into predefined columns of the variants table in the database. In addition, it can store the whole INFO field of each variant into a separate info column so that no information gets lost on the way from the VCF input to the GEMINI database, and that’s the optional behavior we have been asking for. To keep things compact, however, this column stores the INFO content in compressed form, which is not readily accessible.

That’s why it is necessary to extract non-standard INFO field elements explicitly if they are supposed to be used for filtering and querying.

hands_on Hands-on: Making variant call statistics accessible

  1. GEMINI annotate tool:
    • param-file “GEMINI database”: output of GEMINI load
    • param-file “Dataset to use as the annotation source”: output of VarScan somatic
    • “Strict variant-identity matching of database and annotation records (VCF format only)”: Yes

      This setting does not really matter here since you have built the GEMINI database from the exact same list of variants that we are now retrieving annotations from and because VarScan somatic does not call multiple alleles at single sites. Matching on variant-identity is the behavior we would like to see though, so we may as well be explicit about it.

    • “Type of information to add to the database variants”: Specific values extracted from matching records in the annotation source (extract)

      • In “1: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: SS
        • “Database column name to use for recording annotations”: somatic_status
        • “What type of data are you trying to extract?”: Integer numbers
        • “If multiple annotations are found for the same variant, store …“: the first annotation found

        This is the recipe for extracting the VarScan-generated SS field and adding it as a new column somatic_status to the GEMINI database.

      • param-repeat “Insert Annotation extraction recipe”
      • In “2: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: GPV
        • “Database column name to use for recording annotations”: germline_p
        • “What type of data are you trying to extract?”: Numbers with decimal precision
        • “If multiple annotations are found for the same variant, store …“: the first annotation found

        This is the recipe for extracting the VarScan-generated GPV field and adding it as a new column germline_p to the GEMINI database.

      • param-repeat “Insert Annotation extraction recipe”
      • In “3: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: SPV
        • “Database column name to use for recording annotations”: somatic_p
        • “What type of data are you trying to extract?”: Numbers with decimal precision
        • “If multiple annotations are found for the same variant, store …“: the first annotation found

        This is the recipe for extracting the VarScan-generated SPV field and adding it as a new column somatic_p to the GEMINI database.

Next, we are going to add additional annotations beyond the ones directly obtainable through GEMINI or from the input VCF dataset. Specifically we want to add:

  • more information from dbSNP

    As part of the database creation process, GEMINI already checks all variants whether they occur in dbSNP and, if so, stores their dbSNP IDs. Hence, we only need to extract some additional information of interest.

  • information from Cancer Hotspots
  • links to the CIViC database
  • information from the Cancer Genome Interpreter (CGI)

hands_on Hands-on: Adding further annotations

  1. GEMINI annotate tool to add further annotations from dbSNP
    • param-file “GEMINI database”: the output of the last GEMINI annotate tool run
    • param-file “Dataset to use as the annotation source”: the imported dbsnp.b147.chr5_12_17.vcf
    • “Strict variant-identity matching of database and annotation records (VCF format only)”: Yes

      dbSNP stores information about specific SNPs observed in human populations and we would like to record if any exact same SNPs are among our variants.

    • “Type of information to add to the database variants”: Specific values extracted from matching records in the annotation source (extract)
      • In “1: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: SAO
        • “Database column name to use for recording annotations”: rs_ss
        • “What type of data are you trying to extract?”: Integer numbers
        • “If multiple annotations are found for the same variant, store …“: the first annotation found

        This recipe extracts the dbSNP SAO field and adds it as rs_ss to the GEMINI database.

      • param-repeat “Insert Annotation extraction recipe”
      • In “2: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: CFL
        • “Database column name to use for recording annotations”: rs_cfl
        • “What type of data are you trying to extract?”: Numbers with decimal precision
        • “If multiple annotations are found for the same variant, store …“: the first annotation found

        This recipe extracts the dbSNP CFL field and adds it as rs_cfl to the GEMINI database.

      • param-repeat “Insert Annotation extraction recipe”
      • In “3: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: ASP
        • “Database column name to use for recording annotations”: rs_asp
        • “What type of data are you trying to extract?”: Integer numbers
        • “If multiple annotations are found for the same variant, store …“: the first annotation found

        This recipe extracts the dbSNP ASP field and adds it as rs_asp to the GEMINI database.

  2. GEMINI annotate tool to add further annotations from cancerhotspots
    • param-file “GEMINI database”: the output of the last GEMINI annotate tool run
    • param-file “Dataset to use as the annotation source”: the imported cancerhotspots_v2.bed
    • “Strict variant-identity matching of database and annotation records (VCF format only)”: Yes (with input in BED format this setting will be ignored)

      For the cancerhotspots data, we are simply going to record the best q-value associated with any cancerhotspots variant overlapping one of our variant sites.

    • “Type of information to add to the database variants”: Specific values extracted from matching records in the annotation source (extract)
      • In “1: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: 5

          The q-values are stored in the fifth column of the BED dataset.

        • “Database column name to use for recording annotations”: hs_qvalue
        • “What type of data are you trying to extract?”: Numbers with decimal precision
        • “If multiple annotations are found for the same variant, store …“: the smallest of the (numeric) values

        This is the recipe for extracting the q-values of overlapping cancerhotspots sites and adding them as hs_qvalue to the GEMINI database.

  3. GEMINI annotate tool to add links to CIViC
    • param-file “GEMINI database”: the output of the last GEMINI annotate tool run
    • param-file “Dataset to use as the annotation source”: the imported CIViC.bed
    • “Strict variant-identity matching of database and annotation records (VCF format only)”: Yes (with input in BED format this setting will be ignored)

      For the CIViC data, we are going to record the hyperlink to any variant found in the CIViC database that overlaps one of our variant sites.

    • “Type of information to add to the database variants”: Specific values extracted from matching records in the annotation source (extract)
      • In “1: Annotation extraction recipe”:
        • “Elements to extract from the annotation source”: 4

          The hyperlinks are stored in the fourth column of the BED dataset.

        • “Database column name to use for recording annotations”: overlapping_civic_url
        • “What type of data are you trying to extract?”: Text (text)
        • “If multiple annotations are found for the same variant, store …“: a comma-separated list of non-redundant (text) values

        This is the recipe for extracting the hyperlinks of overlapping CIViC sites and adding them as a list of overlapping_civic_urls to the GEMINI database.

  4. GEMINI annotate tool to add information from the Cancer Genome Interpreter (CGI)
    • param-file “GEMINI database”: the output of the last GEMINI annotate tool run
    • param-file “Dataset to use as the annotation source”: the imported cgi_variant_positions.bed
    • “Strict variant-identity matching of database and annotation records (VCF format only)”: Yes (with input in BED format this setting will be ignored)

      For the CGI data, we are going to record if any of our variant sites is overlapped by a variant in the CGI biomarkers database.

    • “Type of information to add to the database variants”: Binary indicator (1=found, 0=not found) of whether the variant had any match in the annotation source (boolean)
      • “Database column name to use for recording annotations”: in_cgidb

Reporting selected subsets of variants

Now that we have built our GEMINI database and enriched it with additional annotations, it is time that we explore the wealth of information stored in it. The goal is to produce filtered variant reports that list specific classes (somatic, germline, LOH) of high-quality variants together with their most relevant annotations. The way to achieve this is through GEMINI queries that specify:

  1. filters we want to apply to the variant list stored in the database
  2. the pieces of information about the filtered variants that we would like to retrieve

comment The GEMINI query language

GEMINI queries are extremely flexible, enabling users to express many different ideas to explore the variant space, and as such, the complete query syntax can be a bit overwhelming for beginners.

In this section, we will start off with a rather simple query, then build on it stepwise before trying a really complex query in the next section.

For a more detailed explanation of the query syntax, you should consult the query section of the GEMINI documentation. Since the GEMINI query syntax is built on the SQLite dialect of SQL, the SQLite documentation, in particular, its chapter on SQLite core functions, is another really helpful resource for understanding more sophisticated queries.

Lets start by configuring a simple query to obtain a report of bona fide somatic variants. Our strategy for retrieving them is to:

  1. rely on the somatic status of the variants called by VarScan somatic tool
  2. disregard questionable variants, for which either a non-negligible amount of supporting sequencing reads is also found in the normal tissue data, or which are only supported by a very small fraction of the reads from the tumor sample

hands_on Hands-on: Querying the GEMINI database for somatic variants

  1. GEMINI query tool with:
    • param-file “GEMINI database”: the fully annotated database created in the last GEMINI annotate tool step
    • “Build GEMINI query using”: Basic variant query constructor
      • param-repeat “Insert Genotype filter expression”
        • “Restrictions to apply to genotype values”: gt_alt_freqs.NORMAL <= 0.05 AND gt_alt_freqs.TUMOR >= 0.10

        With this genotype-based filter, we retain only those variants that are supported by less than 5% of the reads of the normal sample, but by more than 10% of the reads of the tumor sample.

      • “Additional constraints expressed in SQL syntax”: somatic_status = 2

        Among the info stored in the GEMINI database is the somatic status VarScan somatic has called for every variant (remember we used GEMINI annotate to add it). With the condition somatic_status = 2 we retain only those variants passing the genotype filter above and considered somatic variants by the variant caller.

      • In “Output format options”
        • “Type of report to generate”: tabular (GEMINI default)
          • “Add a header of column names to the output”: Yes
          • “Set of columns to include in the variant report table”: Custom (report user-specified columns)
            • In “Choose columns to include in the report”:
              • param-check “chrom”
              • param-check “start”
              • param-check “ref”
              • param-check “alt”
            • “Additional columns (comma-separated)”: gene, aa_change, rs_ids, hs_qvalue, cosmic_ids

        Here we specify, which columns (from the variants table of the GEMINI database) we want to have included, in the specified order, in a tabular variant report.

comment How am I supposed to know these column names?

Obviously, you need to know the names of the columns (in the tables of the GEMINI database) to include them in the report, but how are you supposed to know them?

The standard ones (added by GEMINI load when building the database) are listed in the GEMINI documentation. The non-standard columns (the ones you added with GEMINI annotate) have the names you gave them, when you added them.

Alternatively, to get the tables and column names of a specific database listed, you can use GEMINI database info tool like so:

  • “GEMINI database”: the database you want to explore
  • “Information to retrieve from the database”: Names of database tables and their columns

What about more sophisticated filtering?

hands_on Hands-on: More complex filter criteria

  1. GEMINI query tool with the exact same settings as before, but:
    • “Additional constraints expressed in SQL syntax”: somatic_status = 2 AND somatic_p <= 0.05 AND (filter IS NULL OR rs_ids IS NOT NULL) AND rs_cfl != 1 AND rs_asp != 1

    This translates into “variants classified as somatic with a p-value <= 0.05, which haven’t been flagged as likely false-positives or, if so, are listed in dbSNP, but in there, are not flagged as being assembly-dependent or -specific”.

comment SQL keywords

In the condition above, SQL keywords are given in uppercase. This is not a requirement, but it makes it easier to understand the syntax.

You can check whether any cell in a data table is empty with IS NULL, and whether it contains any value with IS NOT NULL. To combine different filter criteria logically, you can use AND and OR, and parentheses to group conditions if required.

If you have followed all steps up to here exactly, running this job should give you a tabular dataset of 43 variants, and with the annotations in the report it is relatively easy to pick out a few interesting ones. Before we focus on the content of the report, however, we could enhance the report format a bit more.

hands_on Hands-on: SQL-based output formatting

  1. GEMINI query tool with the exact same settings as in the last example, but:
    • In “Output format options”
      • “Additional columns (comma-separated)”: type, gt_alt_freqs.TUMOR, gt_alt_freqs.NORMAL, ifnull(nullif(round(max_aaf_all,2),-1.0),0) AS MAF, gene, impact_so, aa_change, ifnull(round(cadd_scaled,2),'.') AS cadd_scaled, round(gerp_bp_score,2) AS gerp_bp, ifnull(round(gerp_element_pval,2),'.') AS gerp_element_pval, ifnull(round(hs_qvalue,2), '.') AS hs_qvalue, in_omim, ifnull(clinvar_sig,'.') AS clinvar_sig, ifnull(clinvar_disease_name,'.') AS clinvar_disease_name, ifnull(rs_ids,'.') AS dbsnp_ids, rs_ss, ifnull(cosmic_ids,'.') AS cosmic_ids, ifnull(overlapping_civic_url,'.') AS overlapping_civic_url, in_cgidb

This last query adds a lot more annotations to the report, and it also demonstrates the use of the AS keyword to rename columns and of some SQLite functions to clean up the output.

question Question

Compare the new report to the previous one to see what has changed.

details Limitations of genotype column queries

In case you are wondering why the above query does not use rounding on the alternate allele frequencies of the samples, i.e., on gt_alt_freqs.TUMOR and gt_alt_freqs.NORMAL, or why it does not rename these columns, that is because it would break the query.

As a general rule, note that all columns in the variants table starting with gt (the genotype columns, calculated from the genotype fields of a VCF dataset) cannot be used like regular SQLite columns, but are parsed by GEMINI separately. That is why you cannot mix them with regular SQLite elements like functions and alias specifications with AS. You may also have noticed that gt_alt_freqs.TUMOR and gt_alt_freqs.NORMAL do not obey the column order specification of the query, but end up as the last columns of the tabular report. This is another artefact of GEMINI’s special treatment of gt columns.

Generating reports of genes affected by variants

As a final step, let us now try to generate a gene-centered report based on the same somatic variants we just selected above.

Such a gene-centered report would include annotations that apply to a whole gene affected by a variant rather than to the variant itself. Examples of such annotations include known synonyms of an affected gene, its NCBI entrez number, the ClinVar phenotype, if any, associated with the gene, a hyperlink to the gene’s page at CIViC.org, etc..

Some of this information comes built-in into every GEMINI database, but it is stored in a separate table called gene_detailed, while all information we used and queried so far was from the variants table.

To access information from the variants and the gene_detailed table in the same query we need to join the two tables. Such an operation is not possible with the Basic query constructor we have used so far, but requires an advanced mode for composing the query.

hands_on Hands-on: Turning query results into gene-centered reports

  1. GEMINI query tool in advanced mode by choosing
    • “Build GEMINI query using”: Advanced query constructor
    • “The query to be issued to the database”: SELECT v.gene, v.chrom, g.synonym, g.hgnc_id, g.entrez_id, g.rvis_pct, v.clinvar_gene_phenotype FROM variants v, gene_detailed g WHERE v.chrom = g.chrom AND v.gene = g.gene AND v.somatic_status = 2 AND v.somatic_p <= 0.05 AND (v.filter IS NULL OR v.rs_ids IS NOT NULL) AND v.rs_cfl != 1 and v.rs_asp != 1 GROUP BY g.gene

    comment Elements of the SQL query

    In this full SQL query, the part between SELECT and FROM states which columns from which database tables we wish to retrieve, while the part between FROM and WHERE specifies the database tables that need to be consulted and gives them simpler aliases (v becomes an alias for the variants table, g for the gene_detailed table), which we can then use everywhere else in the query.

    The part following WHERE are the filter criteria we want to apply. Note that these criteria are almost the same as those we have used in our earlier somatic variants query, but since we are working with two tables instead of just one, we need to state which table the filter columns come from through table prefixes. Thus, somatic_status becomes v.somatic_status, etc.. In addition, we want to report, of course, corresponding information from the two tables, which is ensured by the additional criteria v.chrom = g.chrom and v.gene = g.gene (the SQL terminology for this is: we want to join the variants and the gene_detailed tables on their chrom and gene columns).

    The GROUP BY part, finally, specifies that we want to collapse records affecting the same gene into one.

  • “Genotype filter expression”: gt_alt_freqs.NORMAL <= 0.05 AND gt_alt_freqs.TUMOR >= 0.10

    This remains the same as in the previous somatic variants query.

Adding additional annotations to the gene-centered report

Unfortunately, GEMINI annotate lets you add columns only to the variants table of a GEMINI database, but there is no simple way to enhance the gene_detailed table with additional annotations. That’s why we are going to add such extra annotations now to the tabular gene-centered report using more general-purpose Galaxy tools.

Annotating the tabular gene report produced with GEMINI is a two-step process, in which we first join the report and tabular annotation sources into a larger tabular dataset, from which we then eliminate redundant and unwanted columns, while rearranging the remaining ones.

Step 1 consists of three separate join operations that sequentially pull in the annotations found in the three gene-based tabular datasets that you imported in the Get Data step of this section.

hands_on Hands-on: Join

  1. Join two files tool to add UniProt cancer genes information
    • param-file “1st file”: the GEMINI-generated gene report from the previous step
    • “Column to use from 1st file”: Column: 1
    • param-file “2nd file”: the imported Uniprot_Cancer_Genes dataset
    • “Column to use from 2nd file”: Column: 1
    • “Output lines appearing in”: Both 1st & 2nd file, plus unpairable lines from 1st file. (-a 1)

      If a variant-affected gene is not listed as a Uniprot cancer gene, then, obviously, we still want to have it included in the final report.

    • “First line is a header line”: Yes
    • “Ignore case”: No
    • “Value to put in unpaired (empty) fields”: 0

      If you inspect the Uniprot_Cancer_Genes dataset, you will see that it has two annotation columns - one indicating, using 1 and 0, whether a given gene is a proto-oncogene or not, the other one indicating tumor suppressor genes the same way. For genes missing from this annotation dataset, we want to fill the corresponding two columns of the join result with 0 to indicate the common case that a gene affected by a variant is neither a known proto-oncogene, nor a tumor suppressor gene.

  2. Join two files tool to add CGI biomarkers information
    • param-file “1st file”: the partially annotated dataset from the previous
    • “Column to use from 1st file”: Column: 1
    • param-file “2nd file”: the imported cgi_genes dataset
    • “Column to use from 2nd file”: Column: 1
    • “Output lines appearing in”: Both 1st & 2nd file, plus unpairable lines from 1st file. (-a 1)
    • “First line is a header line”: Yes
    • “Ignore case”: No
    • “Value to put in unpaired (empty) fields”: 0

    Inspect the input and the result dataset to make sure you understand what happened at this step.

  3. Join two files tool to add gene information from CIViC
    • param-file “1st file”: the partially annotated dataset from step 2
    • “Column to use from 1st file”: Column: 1
    • param-file “2nd file”: the imported GeneSummaries dataset
    • “Column to use from 2nd file”: Column: 3

      The gene column in the CIViC gene summaries annotation dataset is not the first one!

    • “Output lines appearing in”: Both 1st & 2nd file, plus unpairable lines from 1st file. (-a 1)
    • “First line is a header line”: Yes
    • “Ignore case”: No
    • “Value to put in unpaired (empty) fields”: .

    Inspect the input and the result dataset to make sure you understand what happened at this step.

If you took a look at all output datasets as suggested, you will have noticed that each of the join operations kept the gene columns from both input datasets. In addition, we had no control over the order, in which columns got added to the report, nor could we exclude columns.

In Step 2 of the report preparation we are going to address all of these issues and rearrange to get a fully annotated gene report.

hands_on Hands-on: Rearrange to get a fully annotated gene report

  1. Column arrange by header name tool configured like this:
    • param-file “file to rearrange”: the final Join result dataset from step 3
    • In “Specify the first few columns by name”
      • In “1: Specify the first few columns by name”
        • “column”: gene
      • param-repeat “Specify the first few columns by name”
      • In “2: Specify the first few columns by name”
        • “column”: chrom
      • param-repeat “Specify the first few columns by name”
      • In “3: Specify the first few columns by name”
        • “column”: synonym
      • param-repeat “Specify the first few columns by name”
      • In “4: Specify the first few columns by name”
        • “column”: hgnc_id
      • param-repeat “Specify the first few columns by name”
      • In “5: Specify the first few columns by name”
        • “column”: entrez_id
      • param-repeat “Specify the first few columns by name”
      • In “6: Specify the first few columns by name”
        • “column”: rvis_pct
      • param-repeat “Specify the first few columns by name”
      • In “7: Specify the first few columns by name”
        • “column”: is_OG
      • param-repeat “Specify the first few columns by name”
      • In “8: Specify the first few columns by name”
        • “column”: is_TS
      • param-repeat “Specify the first few columns by name”
      • In “9: Specify the first few columns by name”
        • “column”: in_cgi_biomarkers
      • param-repeat “Specify the first few columns by name”
      • In “10: Specify the first few columns by name”
        • “column”: clinvar_gene_phenotype
      • param-repeat “Specify the first few columns by name”
      • In “11: Specify the first few columns by name”
        • “column”: gene_civic_url
      • param-repeat “Specify the first few columns by name”
      • In “12: Specify the first few columns by name”
        • “column”: description
    • “Discard unspecified columns”: Yes

comment Alternative tool suggestion

If your Galaxy server does not offer the Column arrange tool, it will almost certainly offer Cut columns from a table tool, which can be used as a drop-in replacement. Instead of column names, however, this tool expects a comma-separated list of column indexes, like c1,c2 for the first and second column, so you will have to first figure out the column numbers of your columns of interest.

Note: Be sure not to confuse the suggested tool with Cut columns from a table (cut), which does not let you change the order of columns!

Conclusion

In addition to merely calling variants, somatic variant calling tries to distinguish somatic mutations, which are private to tumor tissue, from germline mutations, that are shared by tumor and healthy tissue, and loss-of-heterozygosity events, which involve the loss, from tumor tissue, of one of two alleles found at a biallelic site in healthy tissue.

Dedicated somatic variant callers can perform this classification on statistical grounds, but the interpretation of any list of variants (somatic, germline or LOH) also depends crucially on rich genetic and cancer-specific variant and gene annotations.

keypoints Key points

  • Follow best practices for read mapping, quality control and mapped reads postprocessing to minimze false-positive variant calls.
  • Use a dedicated somatic variant caller to call variants and to classify them into somatic, germline and LOH event variants on statistical grounds.
  • Annotations and queries based on variant properties add relevance to variant and gene reports.
  • A framework like GEMINI is very helpful for managing, annotating and querying lists of variants in a flexible way.
  • Prefer public, free annotation sources to foster reproducibility and information sharing.

Useful literature

Further information, including links to documentation and original publications, regarding the tools, analysis techniques and the interpretation of results described in this tutorial can be found here.

congratulations Congratulations on successfully completing this tutorial!