https://github.com/bigdatagenomics/    https://twitter.com/bigdatagenomics/

Variant Annotation Proposal From Mount Sinai

Status: DRAFT

Introduction

Similar to RFC #2, we’d like to propose one way of storing annotations in a format that is amenable to joining against ADAM variant data. The purpose of this RFC is to propose a schema (in the form of AVRO IDL) to represent both a set of called variants and annotations on those variants.

One particular requirement of our approach is to separate several concepts:

  • The notion of what a variant is, which is represented by the ADAMVariant record:

    • Chromosome
    • Location on the chromosome
    • Reference allele
    • Alternate allele

    We believe that by making a minimalist definition of a variant, and using that as a key in the various other structures, a high degree of consistency is achieved with respect to what annotations and genotypes are actually associated with.

  • A set of annotations on a variant, such as dbSNP ID, Clinvar data, or arbitrary properties of a novel variant, represented by ADAMDatabaseVariantAnnotation. In this proposal, it only contains the dbSNP ID, but, in the future, it will contain a field for each new type of annotation.

  • A set of genotypes that use the variant definition (1) to indicate calls on samples, as represented by the ADAMGenotype record.

Schema

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89

protocol ADAM {
import idl "adam.avdl";

record ADAMContig {
  union { null, string } contigName;
  // Do we need the reference filename if we also have URL?
  union { null, string } referenceName;
  union { null, long }   referenceLength = null;

  union { null, string } referenceURL = null;
  union { null, string } referenceMD5 = null;
}

record ADAMVariant {
  union { null, ADAMContig } contig;
  union { null, long }       position;
  array<Base> referenceAlleles;
  array<Base> variantAlleles;
}

enum ADAMGenotypeAllele {
  Ref, Alt, NoCall
}

// This record represents all stats that, inside a VCF, are stored outside of the
// sample but are computed based on the samples.  For instance,  MAPQ0 is an aggregate
// stat computed from all samples and stored inside the INFO line.
record VariantStatistics {
  union { null, double } baseQRankSum;
  union { null, double } clippingRankSum;
  union { null, int } readDepth;
  union { null, boolean } downsampled;
  union { null, double } fisherStrandBiasPValue; // Phred-scaled.
  union { null, float } haplotypeScore;
  union { null, float } inbreedingCoefficient;
  array<int> alleleCountMLE;
  array<int> alleleFrequencyMLE;
  union { null, float } rmsMapQ;
  union { null, int } mapq0reads;
  union { null, float } mqranksum;
  union { null, boolean } usedForNegativeTrainingSet;
  union { null, boolean } usedForPositiveTrainingSet;
  union { null, float } variantQualityByDepth;
  union { null, float } readPositionRankSum;
  union { null, float } vqslod; // log odds ratio of being a true vs false variant under trained gaussian mixture model
  union { null, string } culprit;
}

record ADAMGenotype {
  union { null, ADAMVariant } variant;

  // Variant-level "provenance" data, i.e. data shared amongst all concurrently genotyped samples
  union { null, VariantStatistics } variantStats;

  union { null, boolean } varIsFiltered = null;  // "null" implies no filters were applied
  array <string> varFilters = null;

  // Sample-level data, i.e. data specific to this particular sample
  union { null, string } sampleId = null;

  // Length is equal to the ploidy
  array <ADAMGenotypeAllele> alleles = null;

  union { null, int } referenceReadDepth;
  union { null, int } alternateReadDepth;
  union { null, int } readDepth;

  union { null, string } filter;
  union { null, int } genotypeQuality;

  // In the ADAM world we split multiallelic VCF lines into multiple
  // single-alternate records.  This bit is set if that happened for this
  // record.
  union { null, boolean } splitFromMultiAllelic;

  array<int> genotypeLikelihoods; // Phred-scaled. Always length 3 since we are not multiallelic.

  union { null, boolean } isPhased = null;
  union { null, string }  phaseSetId = null;
  union { null, int }     phaseQuality = null;
}

record ADAMDatabaseVariantAnnotation {
  union { null, ADAMVariant } variant;

  union { null, int } dbsnpId = null;
}

}

Comments