VCF Parsing - IGV is using htsjdk library for parsing and structure the VCF data.

Key Classes used as part of VCF parsing:

1. Inbuilt classes:

• VCFFileReader - reads the vcf file
• VCFCodec - inbuilt class in htsjdk used for reading, parsing vcf files
  • extracts all information like INFO, FORMAT, FILTER fields from file
  • basically converts a vcf file into particular type of objects (VariantContext)
• VariantContext
  • used for representing single variant record and has details like CHROM, POS, REF, ALT, genotypes
  • stores parsed data in structured way
• Genotype - represents genotype information

2. Custom classes - created on top of built-in to handle logic according to requirements for better visualisation.

• VCFVariant - custom wrapper around existing inbuilt VariantContext
  • more of formatting data
  • calculates allele freq, metadata or required data that has to be shown in UI
• VCFGenotype - formats genotype data

For displaying or rendering elements for visualisation

• VariantTrack - render vcf variants
• VariantRenderer - handle graphical representation of variants

Overall flow:

1. Loading vcf file:

• IGV detects .vcf, .vcf.gz,.bcf files and it automatically selects appropriate parser based on extension using AbstractFeatureReader class from htsjdk.
• selects VCFCodec from htsjdk for parsing VCF

2. Parsing vcf file :

• VCFCodec reads the headers (## metalines from vcf) - INFO, FORMAT, FILTER metadata definitions.
• stores headers and metadata in VariantContext object
• each variant line is converted into VariantContext object (which has position, reference data) - reads each variant as a structured VariantContext object.
• each sample has genotypes which are stored inn VCFGenoType
• extracts genotypes using variantContext.getGenotype()

3. Storing parsed data:

• VariantContext wrapped inside VCFVariant.java to parse each row
• each variant is stored in VCFVariant obj which has Allele frequency (AF), genotype calls (GT)
• now VCFVariant objects are stored inside VariantTrack for track-level storage

4. Rendering variants

• VariantTrack calls VariantRenderer to draw variants
• color coding (gray for 0/0, ...)
• hovering over variant displays - AF, DP, GQ

Loading population-level VCF:

• Batch loads variants for efficient RAM usage
• Indexed VCF Files (.vcf.gz, .bcf) Enable Region-Based Loading which is basically preloading variants in specific genome region(few thousand base pairs around current view) without loading entire file.
• when user zooms or scrolls, IGV loads new regions and discards previously loaded data

Individual-specific VCF

• Loads entire genotype data since it focuses on one sample.