format_snps: Re-format SNP data.

In hemstrow/snpR: Whole-Genome Analysis Tools for Use with Single Nucleotide Polymorphism Data

Re-format SNP data.

Description

format_snps re-formats SNP data into a range of different possible formats for use in snpR functions and elsewhere.

Usage

format_snps(
  x,
  output = "snpRdata",
  facets = NULL,
  n_samp = NA,
  interpolate = "bernoulli",
  outfile = FALSE,
  ped = NULL,
  input_format = NULL,
  input_meta_columns = NULL,
  input_mDat = NULL,
  sample.meta = NULL,
  snp.meta = NULL,
  chr.length = NULL,
  ncp = 2,
  ncp.max = 5,
  chr = "chr",
  position = "position",
  phenotype = "phenotype",
  plink_recode_numeric = FALSE,
  verbose = FALSE
)

Arguments

x	snpRdata object or data.frame. Input data, in any of the above listed input formats.
output	Character, default "snpRdata". The desired output format, see description for details.
facets	Character or NULL, default NULL. Facets over which to break up data for some output formats, such as within one file for genepop or across multiple files for vcf, following the format described in Facets_in_snpR.
n_samp	Integer or numeric vector, default NA. For structure or RAFM outputs. How many random loci should be selected? Can either be an integer or a numeric vector of loci to use.
interpolate	Character or FALSE, default "bernoulli". If transforming to "sn" or "pa" format, notes the interpolation method to be used to fill missing data. Options are "bernoulli", "af", "iPCA", or FALSE. See details.
outfile	character vector, default FALSE. If a file path is provided, a copy of the output will be saved to that location. For some output styles, such as genepop, additional lines will be added to the output to allow them to be immediately run on commonly used programs.
ped	data.frame default NULL. Optional argument for the "plink" output format. A six column data frame containing Family ID, Individual ID, Paternal ID, Maternal ID, Sex, and Phenotype and one row per sample. If provided, outputs will contain information contained in ped. See plink documentation for more details.
input_format	Character, default NULL. Format of x, by default a snpRdata object. See description for details.
input_meta_columns	Numeric, default NULL. If x is not a snpRdata object, optionally specifies the number of metadata columns preceding genotypes in x. See details for more information.
input_mDat	Character, default "NN". If x is not a snpRdata object, the coding for missing genotypes in x (typically "NN" or "0000").
sample.meta	data.frame, default NULL. If x is not a snpRdata object, optionally specifies a data.frame containing meta data for each sample. See details for more information.
snp.meta	data.frame, default NULL. If x is not a snpRdata object, optionally specifies a data.frame containing meta data for each SNP. See details for more information.
chr.length	numeric, default NULL. Chromosome lengths, for ms input files. Note that a single value assumes that each chromosome is of equal length whereas a vector of values gives the length for each chromosome in order.
ncp	numeric or NULL, default 2. Number of components to consider for iPCA sn format interpolations of missing data. If null, the optimum number will be estimated, with the maximum specified by ncp.max. This can be very slow.
ncp.max	numeric, default 5. Maximum number of components to check for when determining the optimum number of components to use when interpolating sn data using the iPCA approach.
chr	character, default "chr". Name of column containing chromosome information, for VCF or plink! output.
position	character, default "position". Name of column containing position information, for VCF output.
phenotype	character, default "phenotype". Optional name of column containing phenotype information, for plink! output.
plink_recode_numeric	Logical, default FALSE. If FALSE, all chrs/scaffs will be renamed to numbers. This may be useful in some cases. If this is FALSE, chromosome names will be checked for leading numbers and replaced with a corresponding letter (0 becomes A, 1 becomes B, and so on).
verbose	Logical, default FALSE. If TRUE, some progress updates will be reported.

Details

While this function can accept a few non-snpRdata input formats, it will reformat to a snpRdata object internally. As such, it takes a facets argument that works identically to elsewhere in the package, as described in Facets_in_snpR. This argument is only used for output formats where facets are important, such as the genepop format.

While this function can be used as an alternative to import.snpR.data when the output argument is set to "snpRdata", this is more complicated and not recommended. Instead, it is simpler import.snpR.data or one of the wrappers in snpR_import_wrappers. The option is kept for backwards compatibility and internal use.

If non-snpRdata is supplied, SNP and sample metadata may be provided. SNP metadata may either be provided in the first few columns of x, the number of which is designated by the input_meta_columns argument, or in a data.frame given as via the snp.meta argument. Sample metadata may be provided in a data.frame via the sample.meta argument.

Output format options:

• ac: allele count format, allele counts tabulated for all samples or within populations.

• genepop: genepop format, genotypes stored as four numeric characters (e.g. "0101", "0204"), transposed, and formatted for genepop. Rownames are individual IDs in genepop format, colnames are SNP ids, matching the first metadata column in input.

• structure: STRUCTURE format, two lines per individual: allele calls stored as single character numeric (e.g. "1", "2"). Allele calls per individual stored on two subsequent lines.

• 0000: numeric genotype tab format, genotypes stored as four numeric characters (e.g. "0101", "0204").

• hapmap: Migrate-n hapmap, allele counts tabulated within populations, in migrate-n hapmap format. Since this migrate-n implementation is iffy, this probably shouldn't be used much.

• NN: character genotype tab format, genotypes stored as actual base calls (e.g. "AA", "CT").

• pa: allele presence/absence format, presence or absence of each possible allele at each possible genotype noted. Interpolation possible, with missing data substituted with allele frequency in all samples or each population.

• rafm: RAFM format, two allele calls at each locus stored in subsequent columns, e.g. locus1.1 locus1.2.

• faststructure: fastSTRUCTURE format, identical to STRUCTURE format save with the addition of filler columns proceeding data such that exactly 6 columns proceed data. These columns can be filled with metadata if desired.

• dadi: dadi format SNP data format, requires two columns named "ref" and "anc" with the flanking bases around the SNP, e.g. "ACT" where the middle location is the A/C snp.

• plink: PLINK! binary input format, requires columns named "position" and one matching the name designated with the 'chr' argument, and may contain a column named "cM", "cm", or "morgans", containing linkage group/chr, snp ID, position in bp, and distance in cM in order to create .bim extended map file.

• sn: Single character numeric format. Each genotype will be listed as 0, 1, or 2, corresponding to 0, 1, or 2 minor alleles. Can be interpolated to remove missing data with the 'interpolate' argument.

• sequoia: sequoia format. Each genotype is converted to 0/1/2/ or -9 (for missing values). Requires columns ID, Sex, BirthYear (or instead of BirthYear - BYmin and BYmax), optional column Yearlast in sample metadata for running Sequoia. For more information see sequoia documentation.

• fasta: fasta sequence format.

• vcf: Variant Call Format, a standard format for SNPs and other genomic variants. Genotypes are coded as 0/0, 0/1, 1/1, or ./. (for missing values), with a healthy serving of additional metadata but very little sample metadata.

• genalex: GenAlEx format. If an outfile is requested, the data will be sorted according to any provided facets and written as an '.xlsx' object.

• snpRdata: a snpRdata object.

Note that for the "sn" format, the data can be interpolated to fill missing data points, which is useful for PCA, genomic prediction, tSNE, and other methods. To do so, specify interpolate = "af" to insert the expected number of minor alleles given SNP allele frequency or "bernoulli" to do binomial draws to determine the number of minor alleles at each missing data point, where the probability of drawing a minor allele is equal to the minor allele frequency. The expected number of minor alleles based on the later method is equal to the interpolated value from the former, but the later allows for multiple runs to determine the impact of stochastic draws and is generally preferred and required for some downstream analysis. It is therefore the default. As a slower but more accurate alternative to "af" interpolation, "iPCA" may be selected. This an iterative PCA approach to interpolate based on SNP/SNP covariance via imputePCA. If the ncp argument is not defined, the number of components used for interpolation will be estimated using estim_ncpPCA. In this case, this method is much slower than the other methods, especially for large datasets. Setting an ncp of 2-5 generally results in reasonable interpolations without the time constraint.

Note also that for the plink format, a .bed binary file can be generated. If the "plink" option is selected and an outfile is designated, R will generate a ".sh" shell file with the same name given in the outfile argument. Running this file will create a plink.bed file.

Input formats:

• NULL or snpRdata: snpRdata object, the default.

• NN: SNP genotypes stored as actual base calls (e.g. "AA", "CT").

• 0000: SNP genotypes stored as four numeric characters (e.g. "0101", "0204").

• snp_tab: SNP genotypes stored with genotypes in each cell, but only a single nucleotide noted if homozygote and two nucleotides seperated by a space if heterozygote (e.g. "T", "T G").

• sn: SNP genotypes stored with genotypes in each cell as 0 (homozyogous allele 1), 1 (heterozygous), or 2 (homozyogus allele 2).

• ms: .ms file, as output from the simulation program ms.

Value

A data.frame or snpRdata object with data in the correct format. May also write a file to the specified path.

Author(s)

William Hemstrom

Melissa Jones

Examples

## Not run: 
#import data to a snpRdata object
## get sample meta data
sample_meta <- 
    data.frame(pop = substr(colnames(stickRAW)[-c(1:3)], 1, 3), 
               fam = rep(c("A", "B", "C", "D"), 
                         length = ncol(stickRAW) - 3), 
               stringsAsFactors = FALSE)
format_snps(stickRAW, input_format = "0000", 
            input_meta_columns = 3, 
input_mDat = "0000", sample.meta = sample_meta)

#allele count, separated by the pop facet.
format_snps(stickSNPs, "ac", facets = "pop")

#genepop:
format_snps(stickSNPs, "genepop")

#STRUCTURE, subsetting out 100 random alleles:
format_snps(stickSNPs, "structure", n_samp = 100)

#STRUCTURE, subseting out the first 100 alleles:
format_snps(stickSNPs, "structure", n_samp = 1:100)

#fastSTRUCTURE
format_snps(stickSNPs, "faststructure")

#numeric:
format_snps(stickSNPs, "0000")

#hapmap for migrate-n:
format_snps(stickSNPs, "hapmap", facets = "pop")

#character:
format_snps(stickSNPs, "NN")

#presence/absence, SNP data:
format_snps(stickSNPs, "pa")

#RAFM, taking only 100 random snps and seperating by pop
format_snps(stickSNPs, "rafm", facets = "pop", n_samp = 100)

#dadi
## add ref and anc snp meta data columns to stickSNPs
dat <- as.data.frame(stickSNPs)
snp.meta(dat) <- cbind(ref = "ATA", anc = "ACT", snp.meta(stickSNPs))
format_snps(dat, "dadi", facets = "pop")

#PLINK! format, not run to avoid file creation
format_snps(stickSNPs, "plink", outfile = "plink_out", chr = "chr")


#PLINK! format with provided ped
ped <- data.frame(fam = c(rep(1, 210), rep("FAM2", 210)), ind = 1:420, 
                  mat = 1:420, pat = 1:420, 
                  sex = sample(1:2, 420, replace = TRUE), 
                  pheno = sample(1:2, 420, replace = TRUE))
format_snps(stickSNPs, "plink", outfile = "plink_out", 
            ped = ped, chr = "chr")
#note that a column in the sample metadata containing phenotypic information
#can be provided to the "phenotype" argument if wished.

#Sequoia format
b <- sample.meta(stickSNPs)
b$ID <- 1:nrow(b)
b$Sex <- rep(c("F", "M", "U"), length.out=nrow(b))
b$BirthYear <- round(runif(n = nrow(b), 1,1))
a <- stickSNPs
b$ID <- paste0(sample.meta(a)$pop, sample.meta(a)$fam, sample.meta(a)$.sample.id)
sample.meta(a) <- b
format_snps(x = a, output = "sequoia")
#note: if using the birth year windows BYmin and BYmax and or Yearlast, 
ensure that the column names are not stored as BY.min, BY.max, Year.last for
snpR. 

# VCF format
test <- format_snps(stickSNPs, "vcf", chr = "chr")

# GenAlEx format (write to file to generate a facet-sorted xlsx file)
test <- format_snps(stickSNPs, "genalex")

## End(Not run)

hemstrow/snpR documentation built on July 15, 2024, 7:14 p.m.