Genetic variant calling is a computational process that identifies differences in a DNA sequence compared to a reference genome. These variations, including single nucleotide polymorphisms (SNPs) and insertions or deletions (indels), are the foundation of genetic diversity. The process is applied in clinical diagnostics to find genetic markers for diseases, in personalized medicine to tailor treatments, and in evolutionary biology to understand population histories. Pinpointing where an individual’s genetic code diverges from a standard provides the raw data for many biological and medical investigations.
The Variant Calling Process
The process from a biological sample to a list of genetic variants involves a standardized bioinformatics pipeline. It begins with raw sequencing data in a FASTQ file format, which contains the short DNA sequences, or “reads,” from a sequencing machine. The first step is read alignment, where these reads are mapped to their position on a reference genome using tools like Burrows-Wheeler Aligner (BWA) or Bowtie2. This task produces a Sequence Alignment Map (SAM) file or its compressed binary version, a BAM file.
With reads aligned, the pipeline moves to variant identification. The software scrutinizes the aligned reads at every position in the genome, looking for systematic differences from the reference sequence. For example, if the reference has an “A” at a specific location but many reads from the sample show a “G,” the software flags a potential variant. Algorithms use statistical models to differentiate true biological variants from sequencing errors.
The output is a list of potential variants stored in a Variant Call Format (VCF) file. This initial list is often noisy with false positives from sequencing or alignment errors, so the final stage involves filtering and annotation. Filtering removes low-confidence calls based on quality scores, read depth, and mapping quality. Annotation then adds biological context by linking variants to genes, predicting their impact, and noting their frequency in population databases.
Types of Variant Calling Software
Variant calling software is not a one-size-fits-all solution, as tools are designed for specific biological questions and data types. The most common distinction is between callers for inherited variants (germline) and those for mutations acquired in a subset of cells (somatic). This differentiation affects their underlying algorithms and applications.
Germline variant callers identify inherited variants present in every cell. These tools assume the organism is diploid and determine if a locus is homozygous (both copies are the same) or heterozygous (the copies differ). A prominent example is the Genome Analysis Toolkit’s (GATK) HaplotypeCaller, which performs local re-assembly of reads to improve accuracy. Another tool, FreeBayes, uses a Bayesian statistical framework to determine the most probable genotype.
Somatic variant callers detect mutations that arise in specific cells or tissues, a hallmark of cancer. They are built to handle complexities like tumor purity and clonal heterogeneity. To do this, they analyze a tumor sample and a matched normal sample from the same individual, allowing the software to subtract germline variants and isolate mutations unique to the tumor. Examples include Mutect2, VarScan 2, and Strelka2.
A third category of tools finds structural variants (SVs), which are large-scale changes to the chromosome like long deletions, duplications, or inversions. These changes are missed by callers designed for small SNPs and indels. SVs are challenging to detect from short-read sequencing data, so specialized software like Manta and Lumpy use read-pair mapping and split-read alignments to identify them.
Key Considerations for Software Selection
Choosing the right variant calling software requires balancing several factors against the research goals and sequencing data. The experimental design directly influences which tool will perform best, as no single program excels at all tasks. Key considerations include:
- Sequencing strategy: Data from Whole Genome Sequencing (WGS) presents different challenges than Whole Exome Sequencing (WES) or targeted panels due to variations in coverage depth and uniformity. Some tools are better tuned for the uniform coverage of WGS, while others handle the variable depth of WES more effectively.
- Variant type and frequency: The requirements for finding common, inherited SNPs differ from those for detecting rare somatic mutations in a tumor. The search for large structural variants also necessitates a dedicated SV caller, as explained in the previous section.
- Computational resources: Some software packages, like GATK, are computationally intensive and require substantial CPU time and memory (RAM). Researchers must assess if their high-performance computing infrastructure can support the chosen tool.
- Documentation and support: Well-documented software with clear instructions simplifies analysis and reduces errors. An active user community provides a valuable resource for troubleshooting and learning best practices.
Interpreting and Validating Variant Calls
The output of a variant caller is the beginning of a deeper analytical process. The Variant Call Format (VCF) file must be carefully interpreted to distinguish true genetic variants from artifacts. Understanding the metrics within this file is the first step toward generating high-confidence results.
A VCF file has a tabular format where each row is a variant with specific quality metrics. The QUAL field gives a score reflecting the confidence that a variant exists at that site; a higher score is more reliable. Another metric is Genotype Quality (GQ), which represents the confidence in the assigned genotype for each sample.
Since no algorithm is perfect, the raw output will contain false positives, making a robust filtering strategy necessary. Researchers develop filtering cascades based on VCF quality metrics, setting thresholds for QUAL, GQ, and read depth to remove likely errors. This process is iterative, balancing the risk of removing true variants against allowing false positives to remain.
Variants with biological or clinical implications require experimental confirmation, known as orthogonal validation. Sanger sequencing is a classic method for validating individual SNPs and small indels. For other variant types or low-frequency variants, techniques like digital PCR (dPCR) may be more appropriate, providing the highest level of confidence.