Package 'QUILT'

What output directory to use

What chromosome to run. Should match BAM headers

What method to run (diploid or nipt)

When running imputation, where to start from. The 1-based position x is kept if regionStart <= x <= regionEnd

When running imputation, where to stop

Buffer of region to perform imputation over. So imputation is run form regionStart-buffer to regionEnd+buffer, and reported for regionStart to regionEnd, including the bases of regionStart and regionEnd

Path to file with fetal fraction values, one row per entry, in the same order as the bamlist

Path to file with bam file locations. File is one row per entry, path to bam files. Bam index files should exist in same directory as for each bam, suffixed either .bam.bai or .bai

Path to file with cram file locations. File is one row per entry, path to cram files. cram files are converted to bam files on the fly for parsing into QUILT

Optional, if not specified, sampleNames are taken from the SM tag in the header of the BAM / CRAM file. This argument is the path to file with sampleNames for samples. It is used directly to name samples in the order they appear in the bamlist / cramlist

Path to reference fasta used for making cram files. Only required if cramlist is defined

How many cores to use

How many Gibbs samples to use

How many iterations between first using current haplotypes to update read labels, and using current read labels to get new reference haplotypes, to perform

How many iterations of the seek_its should be burn in. As an example, if n_seek_its is 3 and n_burn_in_seek_its is 2, then only the dosage from the final round is included. If n_seek_its is 4 and n_burn_in_seek_its is 2, then dosages from the last two rounds are used. Default value NA sets n_burn_in_seek_its to n_seek_its minus 1

How many haplotypes to use in the faster Gibbs sampling

How many haplotypes to replace per-iteration after doing the full reference panel imputation

How many top haplotypes to store in each grid site when looking for good matches in the full haplotype reference panel. Large values potentially bring in more haplotype diversity, but risk losing haplotypes that are good matches over shorter distances

When TRUE, output GT entry contains phased genotypes (haplotypes). When FALSE, it is from the genotype posteriors, and masked when the maximum genotype posterior entry is less than 0.9

What fraction of grid sites to use when looking for good matches in the full haplotype reference panel. Smaller values run faster but potentially miss haplotypes

Override the default bgzip-VCF / bgen output name with this given file name. Please note that this does not change the names of inputs or outputs (e.g. RData, plots), so if outputdir is unchanged and if multiple QUILT runs are processing on the same region then they may over-write each others inputs and outputs.

Can be used to name interim and misc results from imputation to save an an RData file. Default NULL means do not save such output

Override the default location for miscellaneous outputs saved in RData format

Optional path to prepared RData file with reference objects. Can be used instead of outputdir to coordinate use of QUILT_prepare_reference and QUILT

If preparing reference as part of running QUILT, whether to save the prepared reference output file. Note that if the reference was already made using QUILT_prepare_reference, this is ignored

What directory to use as temporary directory. If set to NA, use default R tempdir. If possible, use ramdisk, like /dev/shm/

Minimum BQ for a SNP in a read. Also, the algorithm uses bq<=mq, so if mapping quality is less than this, the read isnt used

Whether to use (TRUE) or not use (FALSE) bases in soft clipped portions of reads

Integer number of reference haplotypes to use, set to NA to use all of them

Optional, only needed when using genfile or phasefile. File with positions of where to impute, lining up one-to-one with genfile. File is tab seperated with no header, one row per SNP, with col 1 = chromosome, col 2 = physical position (sorted from smallest to largest), col 3 = reference base, col 4 = alternate base. Bases are capitalized. Example first row: 1<tab>1000<tab>A<tab>G<tab>

Path to gen file with high coverage results. Empty for no genfile. If both genfile and phasefile are given, only phasefile is used, as genfile (unphased genotypes) is derivative to phasefile (phased genotypes). File has a header row with a name for each sample, matching what is found in the bam file. Each subject is then a tab seperated column, with 0 = hom ref, 1 = het, 2 = hom alt and NA indicating missing genotype, with rows corresponding to rows of the posfile. Note therefore this file has one more row than posfile which has no header

Path to phase file with truth phasing results. Empty for no phasefile. Supercedes genfile if both options given. File has a header row with a name for each sample, matching what is found in the bam file. Each subject is then a tab seperated column, with 0 = ref and 1 = alt, separated by a vertical bar |, e.g. 0|0 or 0|1. Note therefore this file has one more row than posfile which has no header.

How much of a difference to allow the reads to make in the forward backward probability calculation. For example, if P(read | state 1)=1 and P(read | state 2)=1e-6, re-scale so that their ratio is this value. This helps prevent any individual read as having too much of an influence on state changes, helping prevent against influence by false positive SNPs

Whether to make some plots of per-sample imputation. Especially nice when truth data. This is pretty slow though so useful more for debugging and understanding and visualizing performance

Whether to make some plots of per-sample imputation looking at how the block Gibbs is performing. This can be extremely slow so use for debugging or visualizing performance on one-off situations not for general runs

whether to be more verbose when running

Parameter that controls how to detect ancestral haplotypes that are shuffled during EM for possible re-setting. If set (not NULL), then recombination rate is calculated around pairs of SNPs in window of twice this value, and those that exceed what should be the maximum (defined by nGen and maxRate) are checked for whether they are shuffled

Do not use reads with an insert size of more than this value

Whether to output read labels after the final Gibbs samplings (i.e. whether reads were assigned to arbitrary labelled haplotype 1 or 2 and the probability)

Whether to store what read labels were used during the Gibbs samplings (i.e. whether reads were assigned to arbitrary labelled haplotype 1 or 2)

Whether to record interim dosages or not

Whether to try and use BX tag in same to indicate that reads come from the same underlying molecule

When using BX tag, at what distance between reads to consider reads with the same BX tag to come from different molecules

Whether to add optimal haplotypes to vcf when phasing information is present, where optimal is imputation done when read label origin is known

When using phasefile with known truth haplotypes, infer truth read labels, and use them to infer the real base quality against the bam recorded base qualities

When set to TRUE, then when using a smaller reference panel size (fewer haplotypes than Ksubset), parameter choices are reset appropriately. When set to FALSE, original values are used, which might crash QUILT

For HLA imputation, the physical position closest to the centre of the gene

The seed that controls random number generation. When NA, not used#

Whether to use QUILT to generate posterior state probabilities as part of QUILT-HLA

What coverage to downsample individual sites to. This ensures no floating point errors at sites with really high coverage

For non-Gibbs full imputation, minimum allowed value in haplotype gl, after normalization. In effect, becomes 1/minGLValue becomes maximum difference allowed between genotype likelihoods

Minimum number of sample reads a sample must have for imputation to proceed. Samples that have fewer reads than this will not be imputed in a given region and all output will be set to missing

Number of generations since founding or mixing. Note that the algorithm is relatively robust to this. Use nGen = 4 * Ne / K if unsure

Path to reference VCF file with haplotypes, matching the reference haplotype and legend file

Path to reference haplotype file in IMPUTE format (file with no header and no rownames, one row per SNP, one column per reference haplotype, space separated, values must be 0 or 1)

Path to reference haplotype legend file in IMPUTE format (file with one row per SNP, and a header including position for the physical position in 1 based coordinates, a0 for the reference allele, and a1 for the alternate allele)

Path to reference sample file (file with header, one must be POP, corresponding to populations that can be specified using reference_populations)

Vector with character populations to include from reference_sample_file e.g. CHB, CHS

Phred scaled likelihood or an error of reference haplotype. Higher means more confidence in reference haplotype genotypes, lower means less confidence

File with one column of samples to exclude from reference samples e.g. in validation, the samples you are imputing

File with regions to exclude from constructing the reference panel. Particularly useful for QUILT_HLA, where you want to exclude SNPs in the HLA genes themselves, so that reads contribute either to the read mapping or state inference. This file is space separated with a header of Name, Chr, Start and End, with Name being the HLA gene name (e.g. HLA-A), Chr being the chromosome (e.g. chr6), and Start and End are the 1-based starts and ends of the genes (i.e. where we don't want to consider SNPs for the Gibbs sampling state inference)

Path to file with genetic map information, a file with 3 white-space delimited entries giving position (1-based), genetic rate map in cM/Mbp, and genetic map in cM. If no file included, rate is based on physical distance and expected rate (expRate)

Integer Maximum number of distinct haplotypes to store in reduced form. Recommended to keep as 2 ** N - 1 where N is an integer greater than 0 i.e. 255, 511, etc

Whether to output a list of sites as a minimal VCF, for example to use with GATK 3 to genotype given sites

If make_fake_vcf_with_sites_list is TRUE, optional desired filename where to output sites VCF

Expected recombination rate in cM/Mb

Maximum recomb rate cM/Mb

Minimum recomb rate cM/Mb

Print extra timing information, i.e. how long sub-processes take, to better understand why things take as long as they do

What iterations to perform block Gibbs sampling for the Gibbs sampler

How many iterations to run the Gibbs sampler for each time it is run (i.e. how many full passes to run the Gibbs sampler over all the reads)

Plot per sample likelihoods i.e. the likelihood as the method progresses through the Gibbs sampling iterations

For testing purposes only

How many neighouring haplotypes to scan up and down at each grid.

Minimun long grids matches

Use msPBWT to select new haplotypes

How many mspbwt indices to build

Use split real GL in hap selection and imputation

Not for general use. If NA will choose version appropriately depending on whether a PBWT flavour is used.

Used for nMaxDH less than or equal to 255. Use R raw format to hold hapMatcherR. Lowers RAM use

When using shard gibbs sampler, whether to check every pair of SNPs, or not

Use eigen library for per haploid full li and stephens pass of full haplotype reference panel

Whether to use common SNPs first for imputation, followed by a round of rare imputation

Allele frequency yhreshold under which SNPs are considered rare, otherwise they are considered common

Whether to make a plot for understanding heuristic performance

Which heuristic to use

If when using mspbwt, use columns of A rather than the whole thing, to speed up this version

If when calculating genProbs, calculate gamma on the fly rather than saving #' @return Results in properly formatted version

Path to file with bam file locations. File is one row per entry, path to bam files. Bam index files should exist in same directory as for each bam, suffixed either .bam.bai or .bai

HLA region to be analyzed, for example A for HLA-A

Path to dictionary file for reference build

For reference packages built after QUILT 1.0.2, this is not used. For older reference packages, this is needed, and is a path to file with gene boundaries. 4 columns, named Name Chr Start End, with respectively gene name (e.g. HLA-A), chromsome (e.g. chr6), and 1 based start and end positions of gene

What output directory to use. Otherwise defaults to current directory

Prefix for output text summary files

How many cores to use

Output directory containing HLA reference material necessary for QUILT HLA

Prepared HLA haplotype reference panel file

Final output RData file path, if desired

Whether to write out final summary text files or not

How many QUILT Gibbs samples to perform

How many seek iterations to use in QUILT Gibbs sampling

When running QUILT Gibbs sampling, what seed to use, optionally

What chromosome, probably chr6 or maybe 6

For QUILT Gibbs sampling, what buffer to include around center of gene

For QUILT Gibbs sampling, do not consider sequence information if the base quality is below this threshold

When reporting results, give results until posterior probability exceeds this value

For imputing states specifically using QUILT, what coverage to downsample individual sites to. This ensures no floating point errors at sites with really high coverage. This is not used in the direct read mapping

What output directory to use

Number of generations since founding or mixing. Note that the algorithm is relatively robust to this. Use nGen = 4 * Ne / K if unsure, where K is number of haplotypes.

Path to file with 1000 Genomes formatted HLA types (see example for format details)

Path to zip file with alignments from IPD-IGMT (see README and example for more details)

Path to reference genome fasta

Path to file with UCSC refseq gene information (see README and example for more details)

When building HLA full reference panel, start of maximal region spanning all HLA genes. The 1-based position x is kept if regionStart <= x <= regionEnd

As above, but end of maximal region spanning all HLA genes

Buffer of region to perform imputation over. So imputation is run form regionStart-buffer to regionEnd+buffer, and reported for regionStart to regionEnd, including the bases of regionStart and regionEnd

File with regions to exclude from constructing the reference panel. Particularly useful for QUILT_HLA, where you want to exclude SNPs in the HLA genes themselves, so that reads contribute either to the read mapping or state inference. This file is space separated with a header of Name, Chr, Start and End, with Name being the HLA gene name (e.g. HLA-A), Chr being the chromosome (e.g. chr6), and Start and End are the 1-based starts and ends of the genes (i.e. where we don't want to consider SNPs for the Gibbs sampling state inference)

Path to file with genetic map information, a file with 3 white-space delimited entries giving position (1-based), genetic rate map in cM/Mbp, and genetic map in cM. If no file included, rate is based on physical distance and expected rate (expRate)

Path to reference haplotype file in IMPUTE format (file with no header and no rownames, one row per SNP, one column per reference haplotype, space separated, values must be 0 or 1)

Path to reference haplotype legend file in IMPUTE format (file with one row per SNP, and a header including position for the physical position in 1 based coordinates, a0 for the reference allele, and a1 for the alternate allele)

Path to reference sample file (file with header, one must be POP, corresponding to populations that can be specified using reference_populations)

File with one column of samples to exclude from final reference panels. These samples can be removed at one of two points, depending on reference_exclude_samples_for_initial_phasing. If reference_exclude_samples_for_initial_phasing is FALSE, then these samples, if present, are used for the initial phasing and allele assignment. If TRUE, then they are removed immediately

See help for reference_exclude_samplelist_file

Character vector of all HLA genes for which IPD-IMGT files are available for

Character vector of HLA genes to prepare for imputation

What chromosome to run (probably chr6 or similar)

Minimum recomb rate cM/Mb

How many cores to use

What output directory to use

What chromosome to run. Should match BAM headers

Number of generations since founding or mixing. Note that the algorithm is relatively robust to this. Use nGen = 4 * Ne / K if unsure

When running imputation, where to start from. The 1-based position x is kept if regionStart <= x <= regionEnd

When running imputation, where to stop

Buffer of region to perform imputation over. So imputation is run form regionStart-buffer to regionEnd+buffer, and reported for regionStart to regionEnd, including the bases of regionStart and regionEnd

Path to output RData file containing prepared haplotypes (has default value that works with QUILT)

Path to reference haplotype file in VCF format (values must be 0 or 1)

Path to reference haplotype file in IMPUTE format (file with no header and no rownames, one row per SNP, one column per reference haplotype, space separated, values must be 0 or 1)

Path to reference haplotype legend file in IMPUTE format (file with one row per SNP, and a header including position for the physical position in 1 based coordinates, a0 for the reference allele, and a1 for the alternate allele)

Path to reference sample file (file with header, one must be POP, corresponding to populations that can be specified using reference_populations)

Vector with character populations to include from reference_sample_file e.g. CHB, CHS

Phred scaled likelihood or an error of reference haplotype. Higher means more confidence in reference haplotype genotypes, lower means less confidence

File with one column of samples to exclude from reference samples e.g. in validation, the samples you are imputing

File with regions to exclude from constructing the reference panel. Particularly useful for QUILT_HLA, where you want to exclude SNPs in the HLA genes themselves, so that reads contribute either to the read mapping or state inference. This file is space separated with a header of Name, Chr, Start and End, with Name being the HLA gene name (e.g. HLA-A), Chr being the chromosome (e.g. chr6), and Start and End are the 1-based starts and ends of the genes (i.e. where we don't want to consider SNPs for the Gibbs sampling state inference)

Path to file with genetic map information, a file with 3 white-space delimited entries giving position (1-based), genetic rate map in cM/Mbp, and genetic map in cM. If no file included, rate is based on physical distance and expected rate (expRate)

Integer Maximum number of distinct haplotypes to store in reduced form. Recommended to keep as 2 ** N - 1 where N is an integer greater than 0 i.e. 255, 511, etc

What directory to use as temporary directory. If set to NA, use default R tempdir. If possible, use ramdisk, like /dev/shm/

Whether to output a list of sites as a minimal VCF, for example to use with GATK 3 to genotype given sites

If make_fake_vcf_with_sites_list is TRUE, optional desired filename where to output sites VCF

Expected recombination rate in cM/Mb

Maximum recomb rate cM/Mb

Minimum recomb rate cM/Mb

Build mspbwt indices to be used in imputation

How many mspbwt indices to build

Not for general use. If NA will choose version appropriately depending on whether a PBWT flavour is used.

Used for nMaxDH less than or equal to 255. Use R raw format to hold hapMatcherR. Lowers RAM use

Whether to use common SNPs first for imputation, followed by a round of rare imputation

Allele frequency threshold under which SNPs are considered rare, otherwise they are considered common.

If when using mspbwt, use columns of A rather than the whole thing, to speed up this version