Welcome to OpenCAP’s documentation!¶

About¶

Welcome to the Open Source CIViC Annotation Pipeline (OpenCAP). This resource is a step-by-step tutorial that allows users to build a custom capture panel that is linked to clinical variant annotations. This capture panel development is accomplished using the publicly available CIViC database. Through the tutorial, we first describe the CIViC database and introduce users to relevant information within CIViC that can be used to build a capture panel. We then go through methods to build a CIViC capture panel using existing variants within the CIViC database. Specifically, variants of interest are identified using custom filters and variant coordinates are queried using CIViC’s public API. Coordinates identified are then used to design custom capture probes that target variants of interest. After custom capture panel development, we provide general guidance on how to prospectively employ capture sequencing reagents on tumor samples. This includes sample preparation, nucleic acid isolation, library preparation, high throughput sequencing, and somatic variant calling. Once variants are identified using automated somatic variants calling and somatic variant refinement, we then show users how to link these variants back to the CIViC database to annotate the tumor sample for clinical relevance. Successful execution of this tutorial will provide users with a unique capture panel customized to the individual’s need with linkage to clinical relevance summaries for all variants within the panel.

OpenCAP is intended for research use only and clinical applications of any panels designed using the SOP would require further panel validation and development in an appropriate setting (e.g., a CLIA certified and CAP accredited laboratory).

CIViC introduction¶

What is CIViC?¶

The Clinical Interpretations of Variants in Cancer (CIViC) database is an open access, open source, community-driven web resource that captures clinically relevant variants in cancer. CIViC is built on evidence statements, whereby each statement summarizes a variant’s potential clinical relevance as described by a publication. These evidence statements are summarized at the variant level and also at the gene level. An example of this hierarchy has been provided:

What is an Assertion?¶

Evidence items can also be used to build Assertions. CIViC assertions aggregate individual evidence items into a single assessment of the clinical relevance of a variant in a specific disease setting. Assertions also allow for incorporation of guidelines (e.g., ACMG, FDA companion tests, drug approvals, AMP variant levels, NCCN guidelines, etc.). An example of a CIViC assertion is shown below:

Getting Started¶

Below we have provided a screencast entitled, CIViC - Getting Started. This screencast covers: * Description of CIViC and its goals * Navigating through CIViC’s core pages * Browsing, searching, and consuming CIViC knowledgebase content

CIViC Resources¶

We have provided a variety of resources to introduce users to the CIVIC database. Please review the following information about CIViC and the CIViC team:

If you have further issues or wish to report a problem, feel free to email the CIViC team at help@civicdb.org

Contributing to CIViC¶

Any user can browse or search existing curated knowledge within the database. However, users must create an account and log-in to contribute new content to CIViC. Different types of contributions can be found under Example Activities on the CIViC help pages. These activities include:

Adding evidence
Contributing to variant or gene summaries
Revising existing CIViC content
Adding assertions
Other curation tasks such as variant coordinate curation

Once a new evidence item or change to existing evidence is submitted to the CIViC database it will become visible (depending on user display preferences). However, the submission will be listed as a “submitted” or “pending” until it is accepted by an editor. CIViC editors must have attained a sufficient degree of relevant education (typically PhD or MD level), must be extensively familiar with the CIViC interface, have a demonstrated track record of successful curation within the database, and must be approved by two existing editorial members. More information on becoming an editor can be found on the Becoming an Editor Help Docs. An example of how a submitted or revised evidence item becomes accepted in CIViC is shown below:

Regardless of curator status, each activity is recorded in the database. Revision history can be viewed for all items within CIViC and personal contributions can be viewed on an individual’s user profile. To promote user activity, CIViC badges can be earned for various curation actions and the Community Leaderboards show the top CIViC contributors, parsed by activity type.

Adding Evidence Items¶

The main curation activity in CIViC involves adding and editing evidence statements. Below we have provided a screencast entitled, Adding CIViC Evidence to walk users through creating an evidence item in CIViC. This screencast covers:

Scanning a publication for curatable details
Signing into CIViC to Add Evidence
Walking through the Add Evidence form
Viewing the submitted evidence

More information on evidence Items can be found on the CIViC Help Pages under Evidence. This guide provides detailed information on evidence statement inputs including: variant origin, evidence types, evidence levels, and evidence trust ratings. Additionally, when users add evidence items, we provide hints and helpful prompts in the right-hand column to assist with evidence submission.

Editing entities in CIViC¶

Any item in the CIViC interface can be edited using the pencil icon. TO DO: insert pencil icon Below we have provided a screencast entitled, Editing entities in CIViC to walk users through editing items (i.e., evidence, variants, genes or assertions) in CIViC. This screencast covers:

Navigating to an entity’s Edit Form
Importance of edit comments
Identifying entities with pending changes
Navigating to an entity’s suggested changes
Reviewing entity revisions

OpenCAP introduction¶

The Open-sourced CIViC Annotation Pipeline (OpenCAP) is a tutorial that provides users with a method to develop a customized capture panel for which the variants contained within the panel are linked to clinical relevance summaries in the CIViC knowledgebase. This tutorial contains four chapters:

Build Custom Capture Panel: In this section we will use the CIViC interface to identify variants of interest for custom capture. The interface will then be used to download variants of interest, associated clinical descriptions, and curated coordinates. We will then use an interactive interface (jupyter notebook) to format the variant coordinates for probe development. The output from this exercise will be a file that is compatible with commercial probe development companies for custom panel development.
Sequence Samples: This section describes the massively parallel sequencing pipeline. We first detail methods for sample procurement and nucleic acid extraction. Subsequently we provide an overview of library preparation, target enrichment, and next-generation sequencing (NGS) on the Illumina platform. We also touch on new platforms for NGS, including PacBio and Oxford NanoPore. This high-level overview gives brief insight into how we employ custom capture reagents on tumor samples to enrich for sequence regions of interest.
Identify Somatic Variants: The completion of the sequencing pipeline results in raw sequence files (e.g., FASTQ), which are text-based files that represent nucleotide sequences for individual reads. In this section, we briefly describe how to align sequencing reads to a reference genome, how to call somatic variants using automated software, and how to refine called variants using semi-automated processes.
Annotate Variants: After defining a putative list of somatic variants associated with the patient’s tumor, this section describes how to link variants back to the CIViC database to annotate the sample for clinical relevance. We again use an interactive jupyter notebook to import somatic variant calls and output an editable report for the user.

Build Custom Capture Panel¶

In this section we will use the CIViC interface to identify variants of interest for custom capture. The interface will then be used to download variants of interest, associated clinical descriptions, and curated coordinates. We will then open an interactive jupyter notebook to reformat the variant coordinates for probe development. The output from this exercise will be a file that is compatible with commercial probe development companies for custom panel development.

We have built a Binder Jupyter Notebook that contains code to pull in CIViC variants derived from the CIViC Search interface and create a list of genomic coordinates that require capture. While you are reading the tutorial, open the link provided below to start the process of loading a Binder Notebook (Note: loading the Jupyter Notebook can take 5-10 minutes):

Build Jupyter Notebook

Identify variants for capture¶

The CIViC database is constantly being updated with new evidence statements and assertions. Therefore, we have provided a real-time query interface that allows users to build a pool of variants required for custom capture. This interface can be accessed by going to the CIViC website, selecting the “SEARCH” button, and navigating to the Variants” tab: `SEARCH-Variants.

To identify variants for capture, users can add conditions (i.e., search criteria) based on 28 predetermined fields in the drop down menu. If multiple conditions are employed, the user has the option to take the union (i.e., match any of the following conditions) or the intersection (i.e., match all of the following conditions). Additionally, after conditions are employed, the user has the option to further filter the selected variants using column headers in the search grid. The user can filter on Gene Name, Variant Name, Variant Group(s), Variant Types(s), or Description. Once the user is satisfied with the existing variants in the search grid, the user can export the data as a comma-separated values (CSV) file using the “Get Data” button in the search grid. This will provide users with information required to build probes for all variants selected. Below we have provided a screenshot showing how to filter on if the variant contains an assertion:

An example file output created from the CIViC interface can be viewed on the GitHub Repository https://raw.githubusercontent.com/griffithlab/civic-panel/master/binder_interactive/Build_Panel/test_create_variants.tsv

Although this screenshot provides one method to create a variant pool, there are many other examples of criteria that can be useful for identifying variants. Below we have provided a few additional examples of fields that might be helpful for building variant prioritization conditions. Each field has an associated description and links to help documents if applicable.

Field	Description	Example	Associated Help Documents
Assertion	Variant is affiliated with clinical assertions that incorporate multiple evidence statements	“Variant is associated with an assertion”	Assertion - TO DO
CIViC Variant Evidence Score	User can indicate minimum threshold for CIViC Variant Evidence Score	“CIViC Variant Evidence Score is above 20”	Variant Evidence Score - https://civicdb.org/help/variants/variant-evidence-score
Description	Variant description must contain a keyword	“Description contains colorectal cancer”	Variant Summary - https://civicdb.org/help/variants/variants-summary
Disease Implicated (Name)	Variant must contain at least one evidence item that is implicated in the desired disease	“Disease Implicated is Melanoma”	Disease Ontology - http://www.disease-ontology.org/
Evidence Items	User can indicate the required number of evidence items with a certain status	“Evidence Items with status accepted is greater than or equal to 5”	Evidence Monitoring - https://civicdb.org/help/getting-started/monitoring
Gene Name	Entrez gene name associated with variant must meet selected criteria	“Gene Name contains TP53”	Gene Name - https://civicdb.org/help/genes/genes-overview
Name	Variant Name must meet designated criteria	“Name does not contain AMPLIFICATION”	Variant Name - https://civicdb.org/help/variants/variants-naming
Pipeline Type	Variant type is associated with sequence ontology ID(s) that can be evaluated on designated pipeline	“Pipeline Type is DNA-based”	Variant Name - TODO
Variant Type(s)	Variant type, which is the assigned sequence ontology ID, must meet designated criteria	“Variant Type(s) does not contain Transcript Amplification”	Variant Type - https://civicdb.org/help/variants/variants-type

Categorize variants based on variant length¶

The CSV file developed using the CIViC Search interface contains the genomic coordinates that encapsulate the variants of interest (i.e., Custom CIViC Variants). Each line in this file represents a single variant that requires probe development. However, before designing probes for these variants, we must further categorize each variant by variant length. This can be accomplished by using CIViC curated coordinates (i.e., variant start position minus variant stop position plus one). If the variant length is less than 250 base pairs, the variant is eligible for hotspot targeting. If the variant is >250 base pairs, the variant requires tiling of the protein coding exons. For variants that require tiling, there are two different types of tiling.

For variants that are large-scale copy number variants (e.g., “AMPLIFICATION”, “LOSS”, “DELETION”, etc.), sparse tiling is appropriate. Sparse tiling requires creating approximately 10 probes spread across all protein coding exons. For variants that are bucket variants, (e.g., “MUTATION”, “FRAMESHIFT MUTATION”, etc.), full tiling is appropriate. Full tiling requires creating overlapping probes across the entire protein coding exon(s).

Hopefully you have already started building the Binder Notebook as recommended at the beginning of this page. If not please select the following link:

Build Jupyter Notebook

Once the Jupiter Notebook is loaded, you can add your Custom CIViC Variants file to the environment using the “Upload” button. See below:

To launch the Jupyter Notebook select the file entitled, “Build Probes Notebook” in the home directory. Selecting this file will direct you to a new tab in the same browser. Each Jupiter Notebook consist of cells which can contain text or code. Running a cell will execute entered functions. A cell can be run by selecting the cell, holding down the “Shift” key and pressing enter. Instructions for how to run the cells are also provided in the notebook. In the Jupyter Notebook cell that contains the python script, ensure that you change the input variant list file name to match the Custom CIViC Variants file that you uploaded to the home directory. The default file name is ‘test_create_variants.tsv’ - See below:

Once you have changed the input file name, hold down the “Shift” key and select “Enter” to process your Custom CIViC variants. Once the code is done, two new files will appear in the home directory:

INPUT_custom_CIViC_variants.txt = coordinates for all probes required to capture variants of interest without annotation
REFERENCE_custom_CIViC_variants.txt = list of all probes required to capture variants of interest with annotation (gene name, probe id, type of tiling)

You can download these files to your local computer by checking the box next to the file of interest and selecting the “Download” button:

The file entitled “INPUT_custom_CIViC_variants.txt” will serve as an example input file, suitable for IDT probe design. This file should be a tab separated text file whereby each row represents a genomic region that requires coverage. An example file is shown below:

Build custom capture panel¶

After generating the INPUT_custom_CIViC_variants.txt file. You can access custom probe software provided by commercial entities for reagent development. Some of these entities include:

Integrated DNA Technologies (IDT) Custom Probes

Twist Biosciences Oligo Pools

NimbleGen SeqCap

We will demonstrate custom capture panel development using the IDT Target Capture Probe Design & Ordering Tool. First, under “Input Format”, select the “Coordinates (BED)” option. Next, select the “Upload File” option and click on the upload human genomic coordinates button. Upload the file that was prepared using the CIViC interface (INPUT_custom_CIViC_variants.txt).

We also recommend looking at the Design Parameters to ensure proper capture design. Ensure that the following parameters are selected:

Target species = “Human (Feb. 2009 GRCh37/hg19)”
Target Definition = “Full Region”
Probe Length = 120 basepairs
Probe Tiling Density = 1X

Successful upload of the text file should look like this:

Once the files are successfully uploaded, select the “Continue” button to develop the reagent. Of note, you must be logged into the interface to continue with this process. The next steps include reviewing the design, ordering probes, ordering buffers, and ordering blocking oligos. Once the panel design has been reviewed, you can purchase the reagents through the IDT interface.

Sequence Samples¶

This section describes the massively parallel sequencing pipeline. We first detail methods for sample procurement and nucleic acid extraction. Subsequently we provide an overview of library preparation, target enrichment, and next-generation sequencing (NGS). We also touch on new methods for NGS, which includes describing PacBio and NanoPore Sequencing. This high-level overview gives brief insight into how we employ custom capture reagents on tumor samples to enrich for variants of interest. Note, there are many variations of the following sequencing pipeline that may be appropriate for the individual researcher or clinical use case. This section is meant to provide a general overview of a typical pipeline.

Sample procurement¶

For the analysis described here, samples must be derived from a germline tissue (normal sample) and a diseased tissue (tumor sample). Procuring samples from these two sample types requires consideration of the malignancy:

Liquid Cancers: A cancer that begins in blood-forming tissue, such as the bone marrow, or in the cells of the immune system (e.g., leukemia, multiple myeloma, and some lymphomas)
Solid Cancers: An abnormal mass that does not contain cysts or liquid areas (e.g., sarcomas, carcinomas, and some lymphomas)

It is important to note that blood samples may not be suitable as the normal samples for solid cancers if the tumor is metastatic with high circulating tumor cells or circulating tumor DNA. In these cases, buccal swabs or skin biopsies may be better for tumor-normal comparisons.

Sample storage¶

Once samples are procured, they are typically preserved as fresh-frozen (FF) or formalin-fixed paraffin-embedded (FFPE) specimens.

Fresh Frozen: As soon as samples are obtained, fresh frozen preparation requires exposing the sample to liquid nitrogen as quickly as possible. Samples must subsequently be stored in −80°C freezers until extraction.
Formalin-fixed paraffin-embedded: FFPE samples can be prepared using a variety of available kits (e.g., QIAamp DNA FFPE Tissue Kit, MagMAX™ FFPE DNA/RNA Ultra Kit, Quick-DNA/RNA FFPE Miniprep Kit, etc.).

Nucleic acid generation¶

If samples are stored as FFPE blocks, they require FFPE DNA extraction. This can be accomplished using commercially available kits. In general, these kits require paraffin removal and tissue rehydration, tissue digestion, mild reversal of cross-linkage, and nucleic acid purification. If samples are stored as fresh-frozen tissue blocks, they only require nucleic acid purification.

Nucleic acid purification requires cell lysis, binding of nucleic acid, washing off non nucleic acid material, drying of nucleic acid, and elution into a buffer. There are many commercially available kits that can perform nucleic acid purification. These steps can also be automated using commercially available equipment (e.g., QIAsymphony® SP, NUCLISENS® easyMAG®, etc.). Below we describe each step in detail:

Lyse: Tissue samples are typically stored as whole cells. The lysis step is used to disrupt the cellular membrane to expose the nucleic acid. Lysis buffers typically comprises a chaotropic agent, which breaks the hydrogen bond network between water molecules and optionally a surfactant to lower surface tension between membrane components and nucleic acid-containing solution. Some chaotropic agents can include: guanidium thiocyanate or magnesium chloride. Some surfactants can include: Triton-X-100, or sodium dodecyl sulfate.

Bind: After nucleic acid has been suspended in solution, it can be reversibly bound to a positively charged material for purification. These materials can include magnetic particles, columns, filters, silica beads, or organic solvent-based methods.

Wash: Once the nucleic acid is bound to a positively charged material, remaining substances in the lysate are washed from solution. A washing solution does not disrupt the covalent bond between the nucleic acid and the positively charged material used for purification.

Dry: To ensure proper elution, bound nucleic acid typically needs to be completely devoid of all liquid. To avoid degradation, alcohols can be used to expedite the drying step.

Elute: Elution buffers are solvents that displace the nucleic acid from the positively charged material used for purification. Elution buffers can include: 10 mM Tris at pH 8-9, Warmed MilliQ (60 oC), or 1X TE.

Cleanup: Elutions can be optionally treated with either RNAse (RNase ONE™ Ribonuclease, RNase A, etc.) or DNAse (e.g., DNase I <https://www.neb.com/protocols/0001/01/01/a-typical-dnase-i-reaction-protocol-m0303>, Baseline-ZERO™ DNase, etc.) to eliminate nucleic acid that is not being used in downstream analysis.

Quality check: After the nucleic acid generation step, it is recommended to assess the quantity and quality of the final elution. This can be accomplished using spectrophotometry and/or electropherograms.

Spectrophotometry measures a substance’s ability to absorb a specific wavelength, which in turn is a proxy for concentration and purity. First, the sample is exposed to an ultraviolet light at a wavelength of 260 nanometres (nm) and the DNA and RNA in the sample will absorb a relative amount of the light that is proportional to the concentration. Next a photo-detector measures the light that passes through the sample (i.e., not absorbed), which allows you to calculate the quantification of DNA/RNA in the sample. Nucleic acid quantification using spectrophotometry relies on the Beer–Lambert law.

Electropherograms measure the nucleic acid concentration and size using a fluorescent spectrum. The Agilent 2100 Bioanalyzer evaluates the sample using a spectrum of fluorescents. The Qubit 4 Fluorometer utilizes fluorescent dyes that are specific to the target of interest.

Library construction¶

In advance of next-generation sequencing (NGS), construction of sequencing libraries is first required. This typically requires genomic fragmentation, ligation to custom linkers called adapters, and polymerase chain reaction (PCR) amplification.

Genome fragmentation involves breaking the DNA into smaller pieces using physical or chemical means.

Physical fragmentation methods include sonication, nebulization or enzymatic reactions.

Chemical fragmentation relies on hydroxyl radicals to break DNA into fragments, which can accommodate more material, but can induce false positives through novel mutations or transversion artifacts.

Adaptors are chemically synthesized double stranded DNA molecules that make sequencing reactions possible. Adaptors are ligated to DNA fragments and may include sequences to allow binding to a flowcell, sequencing primer sites, sample indexes, unique molecular identifier (UMI) sequences, etc.

PCR amplification is a method to make many copies of a specific DNA segment. PCR requires first denaturing dsDNA to create ssDNA using heat, binding of targeted primers to ssDNA fragments, and elongation of ssDNA to create a copied dsDNA. Amplification is typically performed at multiple steps in the sequencing pipeline.

Target enrichment strategies¶

Target enrichment strategies are used to generate a specific collection of DNA fragments for sequencing. These enrichment strategies are often performed on the constructed sequence library or incorporated into a library construction step.

Hybridization Capture¶

Hybridization capture requires designing specific primers that bind to regions of interest and isolating these bound DNA fragments using chemistry (e.g., use of strepavidin Beads in combination with biotinylated DNA). Genomic DNA that is not bound to the capture probes will be washed away. The remaining DNA, which is enriched for regions of interest, is amplified using PCR and sequenced. Reagents that use hybridization capture include: Swift BioSciences, IDT, Agilent, among others. The process for hybridization capture is described below:

Amplicon Enrichment¶

Amplicon enrichment uses a slightly different strategy for enrichment of regions of interest. Instead of hybridization based capture, regions of interest are amplified by PCR using sets of primer sequences designed to target regions of interest. Reagents that use amplicon sequencing include: QIAGEN, Illumina, and others. An example of the process of amplicon enrichment is shown below:

Unique Molecular Identifiers¶

Unique molecular identifiers (UMIs) are short sequences or molecular tags that can be added to each read during library preparation. Typically, these molecular identifiers are added prior to amplification so that they tag individual DNA molecules observed in the sample. This allows the individual to assign all amplification products to a single originating DNA molecule after sequencing. Through a process of consensus read formation, individual sequencing-related errors can be discounted, decreasing the effective error-rate of sequencing. UMI-based sequencing can take on many forms, each unique to the individual library preparation. An example of the single molecule molecular inversion probe approach is provided below:

Other considerations¶

Of note, for evaluation of RNA, total RNA must be subjected to reverse transcriptase treatment (e.g., ProtoScript® II Reverse Transcriptase, SuperScript™ III Reverse Transcriptase) to generate cDNA prior to library preparation.

High throughput sequencing¶

Next-generation sequencing¶

Sequencing is the final step in data production part of a genomic analysis pipeline. The most commonly used sequencing technique is so-called next-generation sequencing (NGS) or high-throughput sequencing, which evaluates millions of sequences in parallel to dramatically reduce time and cost of the analysis. There are at least two popular platforms (in use clinically) that harness the power of next-generation sequencing to efficiently sequence tumor samples:

Illumina sequencing anneals individual reads to a bead or plate using DNA adaptors and the molecule is amplified through PCR. Amplified reads are sequenced by individually adding single fluorescently tagged and blocked-nucleotides to the complementary DNA sequence and exposing the nucleotide to light to produce a characteristic fluorescence. These blocked-nucleotides can then be un-blocked to allow for an additional base to bind and the process repeated until the whole complementary sequence is elucidated. This platform has a high accuracy rate and can evaluate 50-300 base-pairs per read, and very high-throughput runs producing millions to billions of reads. Each run takes approximately 2-3 days to complete for as little as $1,000 per 30x whole genome sample.

ThermoFisher ION Torrent evaluates hydrogen atoms emitted during polymerization of base pairs, which can be measured as a variation in the solution’s pH. This method has a low error rate for substitutions and point mutations and it is relatively inexpensive with a fast turn-around for data production (2-7 hours per run), however, the platform has higher error rates for insertions and deletions, it cannot read long chains of mononucleotides, and it cannot currently match the throughput of the Illumina sequencing platform.

Third generation sequencing¶

Third Generation Sequencing Platforms: PacBio and NanoPore are considered third generation sequencing technologies that can sequence longer reads at a reduced cost to address the existing problems associated with NGS.

PacBio utilizes hairpin adaptors to create a loop of DNA that can be fed through an immobilized polymerase to add complementary base pairs. As each nucleotide is held in the detection volume by the polymerase, a light pulse identifies the base. This platform requires high quality intact DNA with highly controlled fragmentation and can read strands up to 1Mb in length.

Oxford NanoPore Sequencing utilizes biological transmembrane proteins that translocalize DNA. Measurement of changes in electoral conductivity as the DNA passes through the pore elucidates sequence reads. This platform can evaluate variable length reads and is inexpensive relative to other technologies. Specifically, the MinION device is completely portable, commercially available and can evaluate 20-100MB per run. The tradeoff is its low fidelity rate of only ~85%.

Identify somatic variants¶

The completion of the sequencing pipeline results in raw FASTA files, which are text-based files that represent nucleotide sequencings. In this section, we describe how to align sequencing reads to reference genome, how to call somatic variants using automated software, and how to refine variants using semi-automated processes.

This section was made possible by the wonderful information provided by the Precision Medicine Bioinformatic Course developed in the Griffith Lab.

Outputs from Sequencing Pipeline¶

1) FASTA Format¶

During library preparation and target enrichment, read strands are generated as input for sequencing platforms. These read strands are digitally read by sequencing machines and printed to a FASTA file. If your sequencing machine parameter included paired-end reads, whereby each read was read twice, then you will have two FASTA files per sample. These raw files have a consistent format that can be easily read by aligners. Each line a FASTA file includes a header and a sequence:

> GAPDH_204s.100.1 AATTAGGAGCGATTTGAGATTGCCCCCGATTTATTGACCCGTTTAGCC

> HAPTB_204s.100.1 AAGGCGTGAGAAAGTGCCCGTGGGTAGTGCGGGAGTGGGATGGTAGCC

Raw FASTA files will include any indexes, linkers, or unique-molecular identifiers (UMIs) that were employed in library preparation or hybridization capture. These sequences will need to be trimmed from the raw sequencing reads prior to alignment. Often this trimming process will be performed by software provided by the commercial entity associated with the instrument being used.

As FASTA files are processed, read strands in FASTA files are annotated with additional information including alignment location, quality, strand, etc.

2) FASTQ Files¶

In addition to FASTA files, sequencing runs often produce FASTQ files, which provides a high level overview of sequencing quality. FASTQ files also have a consistent format that can be easily read by aligners. Each line a FASTQ file includes a header, a sequence, a separator, and quality scores.

Read example 1:

> GAPDH_204s.100.1
AATTAGGAGCGATTTGAGATTGCCCCCGATTTATTGACCCGTTTAGCC
+
!``*((_*>*+())))>>>>>***+1.(%%%%%^&****#)CCCCC65

Read example 2:

> HAPTB_204s.100.1
AAGGCGTGAGAAAGTGCCCGTGGGTAGTGCGGGAGTGGGATGGTAGCC
+
!``*(((*(*+()))>>>>>.%%%%^&**#)C65***+()))>>>>>.%

Quality scores are based on the Phred scale and are enclosed using ASCII Annotation characters (for brevity). Each score is calculated differently depending on the technology/instrument used for sequencing.

3) Pre-Alignment QC¶

FASTQ files can be used to generate FastQC Reports. These reports show basic statistics about sequencing (total reads, total poor quality reads, sequence length, GC content, etc.) and provide graphs that give the user a feel for sequencing quality. An example of this type of report is shown below:

..image:: images/FastQC_Report.png

Generating pre-alignment QC can be accomplished following the commands on the PreAlignment QC page provided by the Precision Medicine Bioinformatic Course.

Alignment Strategies¶

The Reference Genome¶

The reference genome approximates the complete representation of the human genetic sequence for the 4 billion base pairs in human DNA. Using a representative assembly prevents the need to build an assembly each time a genome is sequenced, however, there are intrinsic flaws to this approach. Specifically, due to single nucleotide polymorphisms (SNPs) intrinsic to an individual, the reference genome does not perfectly match any one individual. Further, repetitive elements (duplications, inverted repeats, tandem repeats), the reference is often incomplete or incorrect. Therefore, new genome assemblies are constantly being built to improve our ability to resolve the true human genome sequence. Most recently, GRCh37 was published in 2009 and GRCh38 was published in 2013. A summary of genome releases has been provided by UCSC.

Currently, the CIViC database supports variants from NCBI36 (hg18), GRCh37 (hg19), and GTCh38 (hg20), however most variants are associated with reference build GRCh37 (hg19). Therefore, we recommend that for the pipeline, the alignment strategy should use GRCh37 (hg19).

Alignment Algorithms¶

Alignment can be performed using various alignment software. Generally speaking, alignment strategies can either optimize accuracy or processing time.

Optimal solutions include either Spith-Waterman or Needleman-Wunsch alignment strategies. These algorithms are computationally expensive and process read strands slowly.

Fast solutions include hash-based solutions such as Burrows-Wheeler transformation. These algorithms create shortcuts to reduce alignment time.

The input for alignment software is the FASTA files and the output from alignment is a Sequence Alignment Map (SAM) or Binary Alignment Map (BAM). Typically, these software also produce alignment QC, which includes information about mapped reads, coverage, etc.

Sequencing alignment can be accomplished following the commands on the Alignment page provided by the Precision Medicine Bioinformatic Course.

Germline Variant Analysis¶

The next step in the sequencing pipeline is to use paired tumor and normal alignments for germline and somatic variant calling. Germline variant calling consists of identifying single nucleotide polymorphisms (SNPs), insertions/deletions (indels), and structural variants (SVs) that are intrinsic to the normal tissue. Somatic variant calling is a similar process, but it requires the variant to be exclusively observed in the tumor tissue and not present in the germline (normal) tissue. Below we describe automated methods for bother germline and somatic variant calling.

Germline Variant Analysis¶

Germline variant calling can be performed using a variety of software. Typically, our lab uses GATK (genome analysis tool kit) for initial germline calling and variant filtering.

The optimal method for germline variant calling using GATK Haplotype Caller, which considers all SNPs, Indels, and SVs together by creating a local de novo assembly. Although this method is computationally intensive, it improves overall variant calling accuracy by eliminating many false positives.

Germline variant calling can be accomplished following the commands on the Germline SNV and Indel Calling page provided by the Precision Medicine Bioinformatic Course.

Somatic Variant Analysis¶

In addition to germline variant calling, somatic variant calling can be performed identifying differences that are intrinsic to the tumor sample and not observed in the matched normal samples. Somatic variant calling requires looking for single nucleotide variants (SNVs), insertions and deletions (indels), copy number variants (CNVs), structural variants (SVs), and loss of heterozygosity (LOH). These different types of variants can be identified by using various software. Here we will go through each of these automated variant callers to describe the types of variants each caller can identify and subsequent strengths and weaknesses.

Somatic SNV/InDel Calling¶

Algorithms that perform Somatic SNV/Indel Calling include: VarScan, Strelka, and MuTect. It is recommended that aligned bam files are evaluated by multiple different variant callers and subsequently, filtering can be employed by identifying variants that were observed by multiple different callers.

Varscan is a platform-independent mutation caller for targeted, exome, and whole-genome resequencing data and employs a robust heuristic/statistic approach to call variants that meet desired thresholds for read depth, base quality, variant allele frequency, and statistical significance.

Strelka calls germline and somatic small variants from mapped sequencing reads and is optimized for rapid clinical analysis of germline variation in small cohorts and somatic variation in tumor/normal sample pairs. Both germline and somatic callers include a final empirical variant rescoring step using a random forest model to reflect numerous features indicative of call reliability which may not be represented in the core variant calling probability model.

MuTect2 is a somatic SNP and indel caller that combines the DREAM challenge-winning somatic genotyping engine of the original MuTect (Cibulskis et al., 2013) with the assembly-based machinery of HaplotypeCaller.

Somatic Variant Calling with these three automated callers can be accomplished following the commands on the Somatic SNV/InDel Calling module provided by the Precision Medicine Bioinformatic Course.

Somatic SV Calling¶

Manta is a structural variant caller maintained by Illumina and optimized for calling somatic variation in tumor/normal pairs. Structural variants are rearrangements in DNA involving a breakpoint(s). Generally speaking structural variants can fall into four categories:

Insertions: a region is inserted into the DNA

Deletions: a region is deleted in the DNA

Inversions: a section of DNA is reversed

Translocations: a section of DNA is remved and re-inserted in a new region

Somatic Structural Variant calling with Manta can be executed by following the commands on the Somatic SV Calling module provided by the Precision Medicine Bioinformatic Course.

Somatic CNV Calling¶

Copy number alterations occur when a section of the genome is duplicated or deleted. This phenomenon has an important role in evolution for development of homologs/paralogs that allow for development of new function with retention of old function (e.g., alpha and beta hemoglobin). However, these events can also play an intrinsic role in disease development. Examples of Copy Number Alterations are shown below:

There are two algorithms that can be used to identify copy number alterations in tumor samples:

copyCat is an R package used for detecting somatic (experiment/control) copy number aberrations. It works by measuring the depth of coverage from a sequencing experiment. For example in a diploid organism such as human, a single copy number deletion should result in aproximately half the depth (number of reads) compared to the control.

CNVkit is a python package for copy number calling specifically designed for hybrid capture and exome sequencing data. During a typical hybrid capture sequencing experiment the probes capture DNA from the sequencing library, however the probes don’t always bind perfectly. This results in not only the “on-target” regions being pulled from the library for later sequencing but “off-target” as well where the probes didn’t perfectly bind and essentially pulled the wrong region. The effect provides very low read coverage across the entire genome which CNVkit takes advantage of to make CN calls.

Copy Number Variant calling with copyCat and CNVkit can be executed by following the commands on the CNV Calling module provided by the Precision Medicine Bioinformatic Course.

Somatic LOH Calling¶

Loss of heterozygosity is a common genetic event that occurs in cancer whereby one allele is lost. In this segment, the tumor sample appears to be homozygous whereas the same section is heterozygous in the matched normal sample. Methods for calculating sections of LOH requires first calculating the variant allele frequencies (VAFs) in the normal sample to find heterozygous germline positions. Subsequently, we run bam-readcounts on the tumor sample at these same genomic loci and determine if the tumor sample shows homo- or heterozygosity. If there is an area of homozygosity in the tumor sample that is heterozygous in the normal sample, this represents a section of LOH.

Methods for identifying and visualizing section of LOH can be found in the Somatic LOH Calling module provided by the Precision Medicine Bioinformatic Course.

Annotate variants¶

After identifying a putative list of somatic variants associated with the patient’s tumor, this section describes how to link variants back to the CIViC database to annotate the sample/variants for clinical relevance. We again use an interactive interface (jupyter notebook) to import somatic variants calls and output a report that can be easily consumed by the user.

Build Binder Docker Image¶

We have built a Binder Jupyter Notebook that contains code to pull in a BED/BED-like file and link variants to clinical relevance annotations for all variants that have CIViC curation. Please open the link provided below to start this process (Note: loading the Jupyter Notebook can take 5-10 minutes):

Build Jupyter Notebook

Docker Image Set-up¶

Once the Jupiter Notebook is loaded, the interface should look as follows:

Annotate Variants Notebook.ipynb = This notebook is an interactive session that allows users to run python scripts. The specific notebook in this section is set up to run Identified_variants_to_annotation.py. To use this notebook, you must change the input variables (input variant list)

Identified_variants_to_annotation.py = Python script that takes in the somatic variant list (see test_annotate_variants.tsv) and the sample name. The script will iterate through each variant and ascertain if the variant is in CIViC. If a somatic variant is in CIViC, the script will pull all information about the variant (variant descriptions, assertions, and evidence items) and create an OpenCAP output report. After running the script, the output report will be created in the same directory as these files.

test_annotate_variants.tsv = BED-like tab-separated file with variant coordinates that require OpenCAP annotation. The file contains five columns: Chromosome, Start, Stop, Ref, Var. Each row provides genomic coordinates to a single somatic variant that was observed during sequencing. Each variant will be evaluated for presence in the CIViC database using OpenCAP.

Run OpenCAP in Binder¶

Using the Binder Docker Image created above, you can run the provided identified_variants_to_annotation.py script. This script will take in variants from the sequencing pipeline and output a document with annotation information.

Before running the Jupyter Notebook code blocks, you must upload your somatic variants to the home directory. This can be accomplished using the “upload” button on the home directory:

You must also change the sample name to match the sample name of the associated somatic variants. This name will be used to label the output files.

Once you have configured the input variables, you can run the command by holding the “shift” key and selecting “enter”.

OpenCAP Output File¶

After you run Identified_variants_to_annotation.py using the Jupyter Notebook, a file will be generated in the home directory. This file name will start with the sample name and will end with “OpenCAP_report.docx”. Select this file to download the OpenCAP report for your sample. The following screenshot shows you how to download these reports:

The report will look something like the following:

For a variant to be included in OpenCAP it must be a perfect match (i.e., chromosome, start, stop, reference, variant). Currently, OpenCAP does not support matching bucket variants (e.g., TP53 - MUTATION) or variants without specific genomic changes (e.g., KRAS - G12*). We hope to improve the pipeline over time to allow for annotation of these variants.

For somatic variants that have a perfect match with a CIViC entry, a “Clinical Variant” section has been created for this variant. For each entry, the annotation has four distinct parts:

Variant Information: For each variant, we list the following:

Gene name - HUGO Nomenclature

Protein coding change - HGVS Nomenclature for variant protein change

Genomic coordinates - HGVS Nomenclature for variant genomic coordinates

ENST ID - Ensembl transcript identification number for representative transcript

ENSG ID - Ensembl gene identification number for gene

Variant Description: If the variant has a description in CIViC, the variant description has been reproduced in this section. The variant description contains a high-level overview of all evidence statements available for this variant.
Associated Assertions: If the variant has associated assertions, then these assertions have been reproduced in this section. Assertions incorporate multiple evidence items to support a single clinical relevance statement. Typically, assertions include information from nationally recognized organizations such as the NCCN, the FDA, and the AMP.
Associated Evidence Items: This section provides an overview of evidence items associated with the variant. Evidence items have been condensed into a grid with three columns. The first column is a single description of the evidence item, the second column provides the evidence item identification number(s) that support(s) the description, and finally, we provide the Pubmed identifications numbers associated with these evidence statements. For an evidence statement to be included in the grid, it must have an evidence level greater than “C” (Case Study) and it must be “Accepted”.

Welcome to OpenCAP’s documentation!¶

About¶

CIViC introduction¶

What is CIViC?¶

What is an Assertion?¶

Getting Started¶

CIViC Resources¶

Contributing to CIViC¶

Adding Evidence Items¶

Editing entities in CIViC¶

OpenCAP introduction¶

Build Custom Capture Panel¶

Identify variants for capture¶

Categorize variants based on variant length¶

Build custom capture panel¶

Sequence Samples¶

Sample procurement¶

Sample storage¶

Nucleic acid generation¶

Library construction¶

Target enrichment strategies¶

Hybridization Capture¶

Amplicon Enrichment¶

Unique Molecular Identifiers¶

Other considerations¶

High throughput sequencing¶

Next-generation sequencing¶

Third generation sequencing¶

Identify somatic variants¶

Outputs from Sequencing Pipeline¶

1) FASTA Format¶

2) FASTQ Files¶

3) Pre-Alignment QC¶

Alignment Strategies¶

The Reference Genome¶

Alignment Algorithms¶

Germline Variant Analysis¶

Germline Variant Analysis¶

Germline Variant Refinement¶

Somatic Variant Analysis¶

Somatic SNV/InDel Calling¶

Somatic SV Calling¶

Somatic CNV Calling¶

Somatic LOH Calling¶

Somatic Variant Refinement¶

Annotate variants¶

Build Binder Docker Image¶

Docker Image Set-up¶

Run OpenCAP in Binder¶

OpenCAP Output File¶

Indices and tables¶