PharmCAT (Pharmacogenomics Clinical Annotation Tool)

PharmCAT (Pharmacogenomics Clinical Annotation Tool) is a bioinformatics tool designed to analyze genetic variants to predict drug response and tailor medical treatment.

The tool operates in two primary phases:

1. Genotype/Haplotype Inference: Processes VCF files (from NGS or genotyping) to identify pharmacogenomic (PGx) genotypes and infer haplotypes (star alleles).
2. Phenotype/Recommendation Prediction: Uses pharmacogene diplotypes to predict PGx phenotypes and reports drug-prescribing recommendations based on:
   • CPIC guidelines
   • ClinPGx-annotated DPWG guidelines
   • ClinPGx-annotated FDA-approved drug labels

PharmCAT (Pharmacogenomics Clinical Annotation Tool) is a bioinformatics tool designed to analyze genetic variants to predict drug response and tailor medical treatment based on an individual's genetic profile.

The tool operates in two distinct phases:

1. Genotype and Haplotype Inference: It processes VCF files (from NGS or genotyping) to identify pharmacogenomic (PGx) genotypes and infer haplotypes (typically star alleles).
2. Phenotype Prediction and Reporting: It uses pharmacogene diplotypes (maternal and paternal star allele combinations) to predict PGx phenotypes. It then reports drug-prescribing recommendations derived from:
   • CPIC guidelines
   • ClinPGx-annotated DPWG guidelines
   • ClinPGx-annotated FDA-approved drug labels

When performing frequency analysis, be aware of how the Source Diplotype column handles ambiguous or multi-variant calls:

1. Ambiguous Diplotypes (OR): If the Named Allele Matcher cannot determine a single diplotype (due to unphased or incomplete data), entries are concatenated with OR (e.g., *4/[*6 + *10] OR *6/[*4 + *10]). These should be treated as unknown for frequency analysis.
2. Multi-variant Genes (AND): For specific genes like DPYD and RYR1 that use two low-function variants, calls are separated with AND (e.g., c.775A>G AND c.1627A>G (*5)). For frequency analysis, all alleles in these entries should be counted.

PharmCAT uses a Named Allele Matcher to process gene data. The system categorizes genes into two groups based on how they are processed:

1. Genes PharmCAT will attempt to match: The Named Allele Matcher searches the provided sample file for locations associated with these genes and attempts to match them to known allele definitions automatically.
2. Genes handled by outside callers: PharmCAT does not perform automatic allele matching for these genes (with the exception of CYP2D6 if research mode is enabled). Instead, you must use external tools to generate calls and then supply those calls to PharmCAT to enable recommendation matching.

Common external tools for these genes include:

• Stargazer or StellarPGx: For CYP2D6 diplotype calls.
• Optitype: For HLA calls.

PharmCAT processes genetic variation using two primary input types:

1. VCF Files: Used to identify pharmacogenomic (PGx) genotypes and infer haplotypes (star alleles). VCF files must meet specific VCF Requirements.
2. Outside Call Files: Used when PharmCAT cannot call certain genes directly from VCF data (e.g., CYP2D6, HLA-A, and HLA-B) or when you want to provide diplotype calls made by an external tool. These files follow the Outside Call Format.

Use Outside Call files if you need to generate guideline recommendations for genes that require specialized calling methods or if you already have verified diplotype data.

While the pharmcat_pipeline script is recommended for most users due to its simplicity, the system is composed of modular components that can be run independently for more control:

• VCF Preprocessor: A standalone component for preparing VCF data.
• Core PharmCAT Tool: The central engine, which is itself composed of multiple modules that can be executed individually.

To use these components directly instead of the unified pipeline, refer to the documentation for Running the VCF Preprocessor or Running PharmCAT. This approach is useful if you are setting up PharmCAT in a custom environment without Docker.

The script can filter data based on guideline sources using the -gs flag. If you do not specify genes manually via -g, the script uses the following default gene sets:

CPIC (Default) ABCG2, CACNA1S, CFTR, CYP2B6, CYP2C9, CYP2C19, CYP3A5, CYP4F2, DPYD, G6PD, IFNL3, NAT2, NUDT15, RYR1, SLCO1B1, TPMT, UGT1A1, VKORC1, CYP2D6, HLA-A, HLA-B, MT-RNR1

DPWG ABCG2, CYP2B6, CYP2C9, CYP2C19, CYP3A4, CYP3A5, DPYD, NUDT15, SLCO1B1, TPMT, UGT1A1, VKORC1, CYP2D6

PharmCAT is composed of three distinct modules that form a pipeline. You can run the full pipeline or execute individual modules by specifying their respective flags.

1. Named Allele Matcher: Takes a VCF file and identifies diplotypes. Output: JSON results.
2. Phenotyper: Takes diplotypes (from the Matcher or an outside call file) and determines phenotype information. Output: JSON results.
3. Reporter: Takes Phenotyper JSON results and generates a human-readable HTML or JSON report with drug annotations.

Pipeline Flow: VCF file $\rightarrow$ Named Allele Matcher $\rightarrow$ Phenotyper (optionally accepts outside call file) $\rightarrow$ Reporter $\rightarrow$ HTML report.