#include <snarl_caller.hpp>

Inheritance diagram for vg::PoissonSupportSnarlCaller:

Classes
struct	PoissonCallInfo

Public Member Functions
	PoissonSupportSnarlCaller (const PathHandleGraph &graph, SnarlManager &snarl_manager, TraversalSupportFinder &support_finder, const algorithms::BinnedDepthIndex &depth_index, bool use_mapq)

virtual	~PoissonSupportSnarlCaller ()

void	set_baseline_error (double small_variant_error, double large_variant_error)
	Set some parameters. More...

void	set_insertion_bias (double insertion_threshold, double small_insertion_bias, double large_insertion_bias)
	These are multipliers applied to the errors if the site has an insertion. More...

virtual pair< vector< int >, unique_ptr< CallInfo > >	genotype (const Snarl &snarl, const vector< SnarlTraversal > &traversals, int ref_trav_idx, int ploidy, const string &ref_path_name, pair< size_t, size_t > ref_range)
	Get the genotype of a site. More...

virtual void	update_vcf_info (const Snarl &snarl, const vector< SnarlTraversal > &traversals, const vector< int > &genotype, const unique_ptr< CallInfo > &call_info, const string &sample_name, vcflib::Variant &variant)
	Update INFO and FORMAT fields of the called variant. More...

virtual void	update_vcf_header (string &header) const
	Define any header fields needed by the above. More...

Public Member Functions inherited from vg::SupportBasedSnarlCaller
	SupportBasedSnarlCaller (const PathHandleGraph &graph, SnarlManager &snarl_manager, TraversalSupportFinder &support_finder)

virtual	~SupportBasedSnarlCaller ()

void	set_min_supports (double min_mad_for_call, double min_support_for_call, double min_site_support)
	Set some of the parameters. More...

TraversalSupportFinder &	get_support_finder () const
	Get the traversal support finder. More...

virtual int	get_min_total_support_for_call () const
	Get the minimum total support for call. More...

virtual function< bool(const SnarlTraversal &, int iteration)>	get_skip_allele_fn () const
	Use min_alt_path_support threshold as cutoff. More...

Public Member Functions inherited from vg::SnarlCaller
virtual	~SnarlCaller ()

Protected Member Functions
double	genotype_likelihood (const vector< int > &genotype, const vector< SnarlTraversal > &traversals, const set< int > &trav_subset, const vector< int > &traversal_sizes, const vector< double > &traversal_mapqs, int ref_trav_idx, double exp_depth, double depth_err, int max_trav_size, int ref_trav_size)

vector< int >	rank_by_support (const vector< Support > &supports)
	Rank supports. More...

Protected Attributes
double	baseline_error_small = 0.005
	Baseline error rate for smaller variants. More...

double	baseline_error_large = 0.01
	Baseline error rate for larger variants. More...

double	insertion_bias_large = 1.

double	insertion_bias_small = 1.

double	insertion_threshold = 5.

size_t	top_k = 20
	Consider up to the top-k traversals (based on support) for genotyping. More...

size_t	top_m = 100

double	depth_padding_factor = 1.
	padding to apply wrt to longest traversal to snarl ranges when looking up binned depth More...

const algorithms::BinnedDepthIndex &	depth_index
	Map path name to <mean, std_err> of depth coverage from the packer. More...

bool	use_mapq
	MAPQ information is available from the packer and we want to use it. More...

Protected Attributes inherited from vg::SupportBasedSnarlCaller
const PathHandleGraph &	graph

SnarlManager &	snarl_manager

TraversalSupportFinder &	support_finder
	Get support from traversals. More...

int	min_total_support_for_call = 2

size_t	min_mad_for_filter = 1

size_t	min_site_depth = 4

double	min_alt_path_support = 0.5

Additional Inherited Members
Static Protected Member Functions inherited from vg::SupportBasedSnarlCaller
static int	get_best_support (const vector< Support > &supports, const vector< int > &skips)
	Get the best support out of a list of supports, ignoring skips. More...

static double	support_val (const Support &support)
	Relic from old code. More...

Detailed Description

Find the genotype of some traversals in a site using read support and a Poisson model based on expected depth. Inspired, in part, by Paragraph, which uses a similar approach for genotyping break points

Constructor & Destructor Documentation

◆ PoissonSupportSnarlCaller()

vg::PoissonSupportSnarlCaller::PoissonSupportSnarlCaller	(	const PathHandleGraph &	graph,
		SnarlManager &	snarl_manager,
		TraversalSupportFinder &	support_finder,
		const algorithms::BinnedDepthIndex &	depth_index,
		bool	use_mapq
	)

◆ ~PoissonSupportSnarlCaller()

vg::PoissonSupportSnarlCaller::~PoissonSupportSnarlCaller ( )

virtual

Member Function Documentation

◆ genotype()

pair< vector< int >, unique_ptr< SnarlCaller::CallInfo > > vg::PoissonSupportSnarlCaller::genotype	(	const Snarl &	snarl,
		const vector< SnarlTraversal > &	traversals,
		int	ref_trav_idx,
		int	ploidy,
		const string &	ref_path_name,
		pair< size_t, size_t >	ref_range
	)

virtual

Get the genotype of a site.

Implements vg::SnarlCaller.

◆ genotype_likelihood()

double vg::PoissonSupportSnarlCaller::genotype_likelihood	(	const vector< int > &	genotype,
		const vector< SnarlTraversal > &	traversals,
		const set< int > &	trav_subset,
		const vector< int > &	traversal_sizes,
		const vector< double > &	traversal_mapqs,
		int	ref_trav_idx,
		double	exp_depth,
		double	depth_err,
		int	max_trav_size,
		int	ref_trav_size
	)

protected

Compute likelihood of genotype as product of poisson probabilities P[allele1] * P[allle2] * P[uncalled alleles] Homozygous alleles are split into two, with half support each The (natural) logoarithm is returned If trav_subset is not empty, traversals outside that set (and genotype) will be ignored to save time

◆ rank_by_support()

vector< int > vg::PoissonSupportSnarlCaller::rank_by_support ( const vector< Support > & supports )

protected

Rank supports.

◆ set_baseline_error()

void vg::PoissonSupportSnarlCaller::set_baseline_error	(	double	small_variant_error,
		double	large_variant_error
	)

Set some parameters.

◆ set_insertion_bias()

void vg::PoissonSupportSnarlCaller::set_insertion_bias	(	double	insertion_threshold,
		double	small_insertion_bias,
		double	large_insertion_bias
	)

These are multipliers applied to the errors if the site has an insertion.

◆ update_vcf_header()

void vg::PoissonSupportSnarlCaller::update_vcf_header ( string & header ) const

virtual

Define any header fields needed by the above.

Implements vg::SnarlCaller.

◆ update_vcf_info()

void vg::PoissonSupportSnarlCaller::update_vcf_info	(	const Snarl &	snarl,
		const vector< SnarlTraversal > &	traversals,
		const vector< int > &	genotype,
		const unique_ptr< CallInfo > &	call_info,
		const string &	sample_name,
		vcflib::Variant &	variant
	)

virtual

Update INFO and FORMAT fields of the called variant.

Reimplemented from vg::SupportBasedSnarlCaller.

Member Data Documentation

◆ baseline_error_large

double vg::PoissonSupportSnarlCaller::baseline_error_large = 0.01

protected

Baseline error rate for larger variants.

◆ baseline_error_small

double vg::PoissonSupportSnarlCaller::baseline_error_small = 0.005

protected

Baseline error rate for smaller variants.

Error rates are different for small and large variants, which depend more on base and mapping qualities respectively. The switch threshold is in TraversalSupportFinder. Error stats from the Packer object get added to these baselines when computing the scores.

◆ depth_index

const algorithms::BinnedDepthIndex& vg::PoissonSupportSnarlCaller::depth_index

protected

Map path name to <mean, std_err> of depth coverage from the packer.

◆ depth_padding_factor

double vg::PoissonSupportSnarlCaller::depth_padding_factor = 1.

protected

padding to apply wrt to longest traversal to snarl ranges when looking up binned depth

◆ insertion_bias_large

double vg::PoissonSupportSnarlCaller::insertion_bias_large = 1.

protected

multiply error by this much in pressence of insertion (after some testing, this does not in fact seem to help much in practice. best just to boost overall error above. hence not in CLI and off by default)

◆ insertion_bias_small

double vg::PoissonSupportSnarlCaller::insertion_bias_small = 1.

protected

◆ insertion_threshold

double vg::PoissonSupportSnarlCaller::insertion_threshold = 5.

protected

a site is an insertion if one (supported)allele is this many times bigger than another unlike above, default comes from call_main.cpp (todo: straighten this out?)

◆ top_k

size_t vg::PoissonSupportSnarlCaller::top_k = 20

protected

Consider up to the top-k traversals (based on support) for genotyping.

◆ top_m

size_t vg::PoissonSupportSnarlCaller::top_m = 100

protected

Consider up to the tom-m secondary traversals (based on support) for each top traversal (so at most top_k * top_m considered)

◆ use_mapq

bool vg::PoissonSupportSnarlCaller::use_mapq

protected

MAPQ information is available from the packer and we want to use it.

The documentation for this class was generated from the following files:

src/snarl_caller.hpp
src/snarl_caller.cpp

Classes

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ PoissonSupportSnarlCaller()

◆ ~PoissonSupportSnarlCaller()

Member Function Documentation

◆ genotype()

◆ genotype_likelihood()

◆ rank_by_support()

◆ set_baseline_error()

◆ set_insertion_bias()

◆ update_vcf_header()

◆ update_vcf_info()

Member Data Documentation

◆ baseline_error_large

◆ baseline_error_small

◆ depth_index

◆ depth_padding_factor

◆ insertion_bias_large

◆ insertion_bias_small

◆ insertion_threshold

◆ top_k

◆ top_m

◆ use_mapq