seqApply: Apply Functions Over Array Margins
In SeqArray: Data management of large-scale whole-genome sequence variant calls

Description Usage Arguments Details Value Author(s) See Also Examples

Returns a vector or list of values obtained by applying a function to margins of genotypes and annotations.

seqApply(gdsfile, var.name, FUN, margin=c("by.variant", "by.sample"),
    as.is=c("none", "list", "integer", "double", "character", "logical", "raw"),
    var.index=c("none", "relative", "absolute"), parallel=FALSE,
    .useraw=FALSE, .progress=FALSE, .list_dup=TRUE, ...)

`gdsfile`	a `SeqVarGDSClass` object
`var.name`	the variable name(s), see details
`FUN`	the function to be applied
`margin`	giving the dimension which the function will be applied over; `margin="by.variant"` by default
`as.is`	returned value: a list, an integer vector, etc; return nothing by default `as.is="none"`; `as.is` can be a `connection` object, or a GDS node `gdsn.class` object; if "unlist" is used, produces a vector which contains all the atomic components, via `unlist(..., recursive=FALSE)`
`var.index`	if `"none"` (by default), call `FUN(x, ...)` without variable index; if `"relative"` or `"absolute"`, add an argument to the user-defined function `FUN` like `FUN(index, x, ...)` where `index` is an index of variant starting from 1 if `margin = "by.variant"`: `"relative"` for indexing in the selection defined by `seqSetFilter`, `"absolute"` for indexing with respect to all data
`parallel`	`FALSE` (serial processing), `TRUE` (multicore processing), numeric value or other value; `parallel` is passed to the argument `cl` in `seqParallel`, see `seqParallel` for more details.
`.useraw`	`TRUE`, force to use RAW instead of INTEGER for genotypes and dosages; `FALSE`, use INTEGER; `NA`, use RAW for small numbers instead of INTEGER if possible, it is needed to detect data type (RAW or INTEGER) in the user-defined function; for genotypes, 0xFF is missing value if RAW is used
`.progress`	if `TRUE`, show progress information
`.list_dup`	internal use only
`...`	optional arguments to `FUN`

The variable name should be "sample.id", "variant.id", "position", "chromosome", "allele", "genotype", "annotation/id", "annotation/qual", "annotation/filter", "annotation/info/VARIABLE_NAME", or "annotation/format/VARIABLE_NAME".

"@genotype", "annotation/info/@VARIABLE_NAME" or "annotation/format/@VARIABLE_NAME" are used to obtain the index associated with these variables.

"$dosage" is also allowed for the dosages of reference allele (integer: 0, 1, 2 and NA for diploid genotypes).

"$dosage_alt" returns a RAW/INTEGER matrix for the dosages of alternative allele without distinguishing different alternative alleles.

"$num_allele" returns an integer vector with the numbers of distinct alleles.

"$ref" returns a character vector of reference alleles

"$alt" returns a character vector of alternative alleles (delimited by comma)

"$chrom_pos" returns characters with the combination of chromosome and position, e.g., "1:1272721". "$chrom_pos_allele" returns characters with the combination of chromosome, position and alleles, e.g., "1:1272721_A_G" (i.e., chr:position_REF_ALT).

The algorithm is highly optimized by blocking the computations to exploit the high-speed memory instead of disk.

A vector, a list of values or none.

Xiuwen Zheng

seqBlockApply, seqSetFilter, seqGetData, seqParallel, seqGetParallel

# the GDS file
(gds.fn <- seqExampleFileName("gds"))

# display
(f <- seqOpen(gds.fn))

# get 'sample.id
(samp.id <- seqGetData(f, "sample.id"))
# "NA06984" "NA06985" "NA06986" ...

# get 'variant.id'
head(variant.id <- seqGetData(f, "variant.id"))


# set sample and variant filters
set.seed(100)
seqSetFilter(f, sample.id=samp.id[c(2,4,6,8,10)],
    variant.id=sample(variant.id, 10))

# read
seqApply(f, "genotype", FUN=print, margin="by.variant")
seqApply(f, "genotype", FUN=print, margin="by.variant", .useraw=TRUE)

seqApply(f, "genotype", FUN=print, margin="by.sample")
seqApply(f, "genotype", FUN=print, margin="by.sample", .useraw=TRUE)


# read multiple variables variant by variant
seqApply(f, c(geno="genotype", phase="phase", rsid="annotation/id",
    DP="annotation/format/DP"), FUN=print, as.is="none")

# get the numbers of alleles per variant
seqApply(f, "allele",
    FUN=function(x) length(unlist(strsplit(x,","))), as.is="integer")

# output to a file
fl <- file("tmp.txt", "wt")
seqApply(f, "genotype", FUN=sum, na.rm=TRUE, as.is=fl)
close(fl)
readLines("tmp.txt")

seqApply(f, "genotype", FUN=sum, na.rm=TRUE, as.is=stdout())
seqApply(f, "genotype", FUN=sum, na.rm=TRUE, as.is="integer")
# should be identical



################################################################
# with an index of variant

seqApply(f, c(geno="genotype", phase="phase", rsid="annotation/id"),
    FUN=function(index, x) { print(index); print(x); index },
    as.is="integer", var.index="relative")
# it is as the same as
which(seqGetFilter(f)$variant.sel)



################################################################
# reset sample and variant filters
seqResetFilter(f)

# calculate the frequency of reference allele,
#   a faster version could be obtained by C coding
af <- seqApply(f, "genotype", FUN=function(x) mean(x==0L, na.rm=TRUE),
    as.is="double")
length(af)
summary(af)



################################################################
# apply the user-defined function sample by sample

# reset sample and variant filters
seqResetFilter(f)
summary(seqApply(f, "genotype", FUN=function(x) { mean(is.na(x)) },
    margin="by.sample", as.is="double"))

# set sample and variant filters
set.seed(100)
seqSetFilter(f, sample.id=samp.id[c(2,4,6,8,10)],
    variant.id=sample(variant.id, 10))

seqApply(f, "genotype", FUN=print, margin="by.variant", as.is="none")

seqApply(f, "genotype", FUN=print, margin="by.sample", as.is="none")

seqApply(f, c(sample.id="sample.id", genotype="genotype"), FUN=print,
    margin="by.sample", as.is="none")


# close the GDS file
seqClose(f)


# delete the temporary file
unlink("tmp.txt")