// Protocol buffers for representing gnomAD-SV v4 data.

syntax = "proto3";

package annonars.gnomad.gnomad_sv4;

import "annonars/gnomad/gnomad_sv2.proto";

// Enumeration for filters.
enum Filter {
    // unknown
    FILTER_UNKNOWN = 0;
    // All filters passed.
    FILTER_PASS = 1;
    // Low-quality variant that did not pass manual review of supporting
    // evidence.
    FILTER_FAIL_MANUAL_REVIEW = 2;
    // Unacceptably high rate of no-call GTs
    FILTER_HIGH_NCR = 3;
    // SVs that are overlapped by over 50% by IGH or MHC regions, these
    // variants are of low confidence.
    FILTER_IGH_MHC_OVERLAP = 4;
    // Deletions under 1 kbp that are uniquely from wham and have SR-only
    // support.
    FILTER_LOWQUAL_WHAM_SR_DEL = 5;
    // Multiallelic site
    FILTER_MULTIALLELIC = 6;
    // SVs that are enriched for non-reference genotypes in outlier samples,
    // likely indicating noisy or unreliable genotypes.
    FILTER_OUTLIER_SAMPLE_ENRICHED = 7;
    // Multiple large CNVs called at the same locus likely indicates unreliable
    // clustering and/or low-quality multiallelic locus.
    FILTER_REDUNDANT_LG_CNV = 8;
    // Likely reference artifact sites that are homozygous alternative in over
    // 99% of the samples.
    FILTER_REFERENCE_ARTIFACT = 9;
    // Variant is unresolved
    FILTER_UNRESOLVED = 10;
    // Generic failure.
    FILTER_FAIL = 11;
}

// The types of SV in gnomad-SV v2.
enum SvType {
    // unknown
    SV_TYPE_UNKNOWN = 0;
    // Breakend.
    SV_TYPE_BND = 1;
    // Copy number variable region.
    SV_TYPE_CNV = 2;
    // Complex variant.
    SV_TYPE_CPX = 3;
    // Translocation.
    SV_TYPE_CTX = 4;
    // Deletion.
    SV_TYPE_DEL = 5;
    // Duplication.
    SV_TYPE_DUP = 6;
    // Insertion.
    SV_TYPE_INS = 7;
    // Inversion.
    SV_TYPE_INV = 8;
}

// Store the relevant allele counts and frequencies in a given sub cohort.
message AlleleCounts {
    // Total number of alleles genotyped (for biallelic sites) or individuals
    // with copy-state estimates (for multiallelic sites).
    int32 ac = 1;
    // Allele frequency (for biallelic sites) or copy-state frequency (for
    // multiallelic sites).
    float af = 2;
    // Number of non-reference alleles observed (for biallelic sites) or
    // individuals at each copy state (for multiallelic sites).
    int32 an = 3;

    // Hemizygous alternate genotype frequency (biallelic sites only).
    float freq_hemialt = 4;
    // Hemizygous reference genotype frequency (biallelic sites only).
    float freq_hemiref = 5;
    // Heterozygous genotype frequency (biallelic sites only).
    float freq_het = 6;
    // Homozygous alternate genotype frequency (biallelic sites only).
    float freq_homalt = 7;
    // Homozygous reference genotype frequency (biallelic sites only).
    float freq_homref = 8;

    // Total number of individuals with complete genotypes (biallelic sites
    // only).
    int32 n_bi_genos = 9;

    // Number of individuals with hemizygous alternate genotypes (biallelic
    // sites only).
    int32 n_hemialt = 10;
    // Number of individuals with hemizygous reference genotypes (biallelic
    // sites only).
    int32 n_hemiref = 11;
    // Number of individuals with heterozygous genotypes (biallelic sites
    // only).
    int32 n_het = 12;
    // Number of individuals with homozygous alternate genotypes (biallelic
    // sites only).
    int32 n_homalt = 13;
    // Number of individuals with homozygous reference genotypes (biallelic
    // sites only).
    int32 n_homref = 14;
}

// Store the allele counts for the given sub cohort and sub cohort factored by sex.
message AlleleCountsBySex {
    // Overall allele counts in the sub cohort.
    AlleleCounts overall = 1;
    // Allele counts in female/XX karyotype individuals of sub cohort.
    AlleleCounts xx = 2;
    // Allele counts in male/XY karyotype individuals of sub cohort.
    AlleleCounts xy = 3;
}

// gnomAD SV population.
enum Population {
    // unknown
    POPULATION_UNKNOWN = 0;
    // African
    POPULATION_AFR = 1;
    // Amish
    POPULATION_AMI = 2;
    // Admixed American
    POPULATION_AMR = 3;
    // Ashkenazi Jewish
    POPULATION_ASJ = 4;
    // East Asian
    POPULATION_EAS = 5;
    // Finnish
    POPULATION_FIN = 6;
    // Middle Eastern
    POPULATION_MID = 7;
    // Non-Finnish European
    POPULATION_NFE = 8;
    // South Asian
    POPULATION_SAS = 9;
    // Other
    POPULATION_OTHER = 10;
}

// Store the allele counts for the given sub cohort in the given population.
message PopulationAlleleCounts {
    // Name of the population.
    Population population = 1;
    // The overall allele counts and the one by sex.
    AlleleCountsBySex counts = 2;
}

// Store the allele counts for the given cohort.
message CohortAlleleCounts {
    // Name of the cohort, empty for global.
    optional string cohort = 1;
    // The overall allele counts and the one by sex.
    AlleleCountsBySex by_sex = 2;
    // Allele counts for each population.
    repeated PopulationAlleleCounts by_population = 3;
}

// Protocol buffer for the gnomAD-SV v4 VCF record.
message Record {
    // Chromosome name.
    string chrom = 1;
    // 1-based start position.
    int32 pos = 2;
    // End position of the structural variant
    optional int32 end = 3;
    // Chromosome of second breakpoint position.
    optional string chrom2 = 4;
    // End coordinate of second breakpoint position.
    optional int32 end2 = 5;
    // Identifier of the record.
    string id = 6;

    // Site-level filters.
    repeated Filter filters = 7;
    // SV Type.
    SvType sv_type = 8;
    // Refined complex type.
    optional annonars.gnomad.gnomad_sv2.CpxType cpx_type = 9;

    // Variant allele counts in the different cohorts and population.
    repeated CohortAlleleCounts allele_counts = 10;
}