#include <graph_caller.hpp>
Public Member Functions | |
| VCFOutputCaller (const string &sample_name) | |
| virtual | ~VCFOutputCaller () |
| virtual string | vcf_header (const PathHandleGraph &graph, const vector< string > &contigs, const vector< size_t > &contig_length_overrides) const |
| Write the vcf header (version and contigs and basic info) More... | |
| bool | add_variant (vcflib::Variant &var) const |
| void | write_variants (ostream &out_stream, const SnarlManager *snarl_manager=nullptr) |
| void | vcf_fixup (vcflib::Variant &var) const |
| Run vcffixup from vcflib. More... | |
| void | set_translation (const unordered_map< nid_t, pair< string, size_t >> *translation) |
| Add a translation map. More... | |
| void | set_nested (bool nested) |
| Assume writing nested snarls is enabled. More... | |
| void | set_allele_merge (double threshold, int64_t min_len) |
| unordered_set< string > | get_output_contigs () const |
| string | prune_header_contigs (const string &header, const unordered_set< string > &keep) const |
Protected Member Functions | |
| void | add_allele_path_to_info (const HandleGraph *graph, vcflib::Variant &v, int allele, const Traversal &trav, bool reversed, bool one_based) const |
| add a traversal to the VCF info field in the format of a GFA W-line or GAF path More... | |
| void | add_allele_path_to_info (vcflib::Variant &v, int allele, const SnarlTraversal &trav, bool reversed, bool one_based) const |
| legacy version of above More... | |
| string | trav_string (const HandleGraph &graph, const SnarlTraversal &trav) const |
| convert a traversal into an allele string More... | |
| bool | merge_similar_alleles (const PathPositionHandleGraph &graph, const vector< SnarlTraversal > &site_traversals, vector< int > &site_genotype, const string &sample_name, vcflib::Variant &out_variant) const |
| bool | emit_variant (const PathPositionHandleGraph &graph, SnarlCaller &snarl_caller, const Snarl &snarl, const vector< SnarlTraversal > &called_traversals, const vector< int > &genotype, int ref_trav_idx, const unique_ptr< SnarlCaller::CallInfo > &call_info, const string &ref_path_name, int ref_offset, bool genotype_snarls, int ploidy, function< string(const vector< SnarlTraversal > &, const vector< int > &, int, int, int)> trav_to_string=nullptr) |
| tuple< int64_t, int64_t, bool, step_handle_t, step_handle_t > | get_ref_interval (const PathPositionHandleGraph &graph, const Snarl &snarl, const string &ref_path_name) const |
| pair< string, int64_t > | get_ref_position (const PathPositionHandleGraph &graph, const Snarl &snarl, const string &ref_path_name, int64_t ref_path_offset) const |
| used for making gaf traversal names More... | |
| void | flatten_common_allele_ends (vcflib::Variant &variant, bool backward, size_t len_override) const |
| string | print_snarl (const HandleGraph *grpah, const handle_t &snarl_start, const handle_t &snarl_end, bool in_brackets=false) const |
| string | print_snarl (const Snarl &snarl, bool in_brackets=false) const |
| legacy version of above More... | |
| string | print_flipped_snarl (const Snarl &snarl, bool in_brackets=false) const |
| The same as above, but print the snarl as if its orientation has been flipped. More... | |
| void | scan_snarl (const string &allele_string, function< void(const string &, Snarl &)> callback) const |
| void | update_nesting_info_tags (const SnarlManager *snarl_manager) |
Static Protected Member Functions | |
| static bool | snarl_traversal_to_handles (const HandleGraph &graph, const SnarlTraversal &trav, Traversal &out_trav) |
| static int64_t | allele_core_length (const vector< string > &alleles) |
Protected Attributes | |
| vcflib::VariantCallFile | output_vcf |
| output vcf More... | |
| string | sample_name |
| Sample name. More... | |
| vector< vector< pair< pair< string, size_t >, string > > > | output_variants |
| size_t | max_uncalled_alleles = 5 |
| print up to this many uncalled alleles when doing ref-genotpes in -a mode More... | |
| const unordered_map< nid_t, pair< string, size_t > > * | translation |
| bool | include_nested |
| double | allele_merge_threshold = 1.0 |
| int64_t | allele_merge_min_len = 0 |
Static Protected Attributes | |
| static const int64_t | max_vcf_line_length = 2000000000 |
Detailed Description
Helper class that vcf writers can inherit from to for some common code to output sorted VCF
Constructor & Destructor Documentation
◆ VCFOutputCaller()
| vg::VCFOutputCaller::VCFOutputCaller | ( | const string & | sample_name | ) |
◆ ~VCFOutputCaller()
|
virtual |
Member Function Documentation
◆ add_allele_path_to_info() [1/2]
|
protected |
add a traversal to the VCF info field in the format of a GFA W-line or GAF path
◆ add_allele_path_to_info() [2/2]
|
protected |
legacy version of above
◆ add_variant()
| bool vg::VCFOutputCaller::add_variant | ( | vcflib::Variant & | var | ) | const |
Add a variant to our buffer Returns false if the variant line length exceeds VCFOutputCaller::max_vcf_line_length
◆ allele_core_length()
|
staticprotected |
The CORE LENGTH of a variant: the length of the longest allele after stripping the prefix and the suffix that every non-"*" allele shares. This is the single definition of "how big is this variant" behind –cluster-min-len in BOTH vg call and vg deconstruct. It is invariant to how much shared flanking context a caller keeps in its allele strings, which is the point: vg call flattens down to an anchor base while vg deconstruct emits the whole snarl interior, so a raw string length answers differently for the same variant. Consequences, all intended:
- the anchor base flatten_common_allele_ends must leave on every indel is a shared prefix, so it is stripped: a 49bp indel measures 49, not 50.
- REF participates, so a pure deletion measures the deleted length. A maximum over ALTs alone measures 1 for a deletion of any size.
- "*" is a marker, not sequence, so it is excluded from both the affixes and the maximum. That also neutralizes flatten_common_allele_ends being a no-op whenever a "*" is present – without -a because min_allele_len becomes 1 and max_flatten_len decrements to 0, and with -a because "*" matches no base at the first offset compared. Either way the un-flattened boundary sequence is common to every real allele, so it is stripped here. Note this measures the SPAN of the variant, not the size of any one event inside it: a haplotype differing from the reference at two bases 59bp apart has a core length of 60.
◆ emit_variant()
|
protected |
print a vcf variant return value is taken from add_variant (see above)
◆ flatten_common_allele_ends()
|
protected |
clean up the alleles to not share common prefixes / suffixes if len_override given, just do that many bases without thinking
◆ get_output_contigs()
| unordered_set< string > vg::VCFOutputCaller::get_output_contigs | ( | ) | const |
The set of reference contigs that actually have a record. Reads the sort keys of the output buffer, so it costs nothing (no decompression) and does not need the snarl tree. Only meaningful once calling is finished and before write_variants() drains the buffer.
◆ get_ref_interval()
|
protected |
get the interval of a snarl from our reference path using the PathPositionHandleGraph interface the bool is true if the snarl's backward on the path first returned value -1 if no traversal found
◆ get_ref_position()
|
protected |
used for making gaf traversal names
◆ merge_similar_alleles()
|
protected |
Merge near-identical called ALT alleles in an already-populated variant. Must run AFTER SnarlCaller::update_vcf_info and after flatten_common_allele_ends, so that the genotyper and the allele-flattening both see the full pre-merge allele set: merging earlier drops the absorbed allele's reads from AD/DP and from the Poisson caller's total_other_support term. Rewrites the allele-indexed fields (alleles/alt, AT, AD, GL, GT, MAD) and records what was merged in the MAT info field. Returns true if anything merged.
◆ print_flipped_snarl()
|
protected |
The same as above, but print the snarl as if its orientation has been flipped.
◆ print_snarl() [1/2]
|
protected |
print a snarl in a consistent form like >3435<12222 if in_brackets set to true, do (>3435<12222) instead (this is only used for nested caller)
◆ print_snarl() [2/2]
|
protected |
legacy version of above
◆ prune_header_contigs()
| string vg::VCFOutputCaller::prune_header_contigs | ( | const string & | header, |
| const unordered_set< string > & | keep | ||
| ) | const |
Remove ##contig lines whose ID is not in keep, leaving every other line alone. A reference contig that produced no record is not worth declaring: with a gref cover most contigs are fragments, and on a human chromosome a third of them carry nothing.
◆ scan_snarl()
|
protected |
◆ set_allele_merge()
| void vg::VCFOutputCaller::set_allele_merge | ( | double | threshold, |
| int64_t | min_len | ||
| ) |
Enable post-genotyping merging of near-identical called ALT alleles, so that a 1/2 call of two effectively-identical alleles collapses to 1/1 with a single ALT. Uses the same similarity metric and the same core-length gate as "vg deconstruct -L/--cluster-min-len" (a length-weighted Jaccard, except that a pure deletion is scored against the site – see weighted_traversal_similarity). The gate is applied to the alleles each tool emits, and those sets differ, so the two can disagree at a given site: similarity is >= threshold to merge, and min_len > 0 restricts merging to sites whose core length reaches min_len bp (see allele_core_length). A threshold of 1.0 (the default) disables merging entirely.
◆ set_nested()
| void vg::VCFOutputCaller::set_nested | ( | bool | nested | ) |
Assume writing nested snarls is enabled.
◆ set_translation()
| void vg::VCFOutputCaller::set_translation | ( | const unordered_map< nid_t, pair< string, size_t >> * | translation | ) |
Add a translation map.
◆ snarl_traversal_to_handles()
|
staticprotected |
Convert a SnarlTraversal to the handle vector the clustering code works on. Returns false (leaving out_trav unspecified) if the traversal cannot be represented: fewer than two visits (the "*" placeholder pushed for a star allele), or a visit carrying a child Snarl rather than a node, which NestedFlowCaller produces via SnarlGraph::embed_snarl. (LegacyCaller expands its children into node visits in top_down_genotype, so it never reaches here.)
◆ trav_string()
|
protected |
convert a traversal into an allele string
◆ update_nesting_info_tags()
|
protected |
◆ vcf_fixup()
| void vg::VCFOutputCaller::vcf_fixup | ( | vcflib::Variant & | var | ) | const |
Run vcffixup from vcflib.
◆ vcf_header()
|
virtual |
Write the vcf header (version and contigs and basic info)
Reimplemented in vg::NestedFlowCaller, vg::FlowCaller, vg::LegacyCaller, and vg::VCFGenotyper.
◆ write_variants()
| void vg::VCFOutputCaller::write_variants | ( | ostream & | out_stream, |
| const SnarlManager * | snarl_manager = nullptr |
||
| ) |
Sort then write variants in the buffer snarl_manager needed if include_nested is true
Member Data Documentation
◆ allele_merge_min_len
|
protected |
◆ allele_merge_threshold
|
protected |
◆ include_nested
|
protected |
◆ max_uncalled_alleles
|
protected |
print up to this many uncalled alleles when doing ref-genotpes in -a mode
◆ max_vcf_line_length
|
staticprotected |
◆ output_variants
|
mutableprotected |
output buffers (1/thread) (for sorting) variants stored as strings (and position key pairs) because vcflib::Variant in-memory struct so huge
◆ output_vcf
|
mutableprotected |
output vcf
◆ sample_name
|
protected |
Sample name.
◆ translation
|
protected |
The documentation for this class was generated from the following files:
- src/graph_caller.hpp
- src/graph_caller.cpp
