R/mc3_mthds.R

In GladiaTOX: R Package for Processing High Content Screening data

#####################################################################
## This program is distributed in the hope that it will be useful, ##
## but WITHOUT ANY WARRANTY; without even the implied warranty of  ##
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the    ##
## GNU General Public License for more details.                    ##
#####################################################################

#-------------------------------------------------------------------------------
# mc3_mthds: List of normalization methods (to be used at level 3)
#-------------------------------------------------------------------------------

#' @name MC3_Methods
#' @title List of level 3 multiple-concentration normalization methods
#'
#' @description
#' \code{mc3_mthds} returns a list of normalization methods to be used
#' during level 3 multiple-concentration processing.
#'
#' @return A list of functions
#'
#' @seealso \code{\link{mc3}}, \code{\link{gtoxMthdLoad}} to query what methods
#' get applied to each aeid
#'
#' @details
#' The functions contained in the list returned by \code{mc3_mthds} take
#' 'aeids' (a numeric vector of aeid values) and returns a list of expressions
#' to be executed in the \code{mc3} (not exported) function environment. The
#' functions are described here for reference purposes, The
#' \code{mc3_mthds} function is not exported, nor is it intended for use.
#'
#' All available methods are described in the Available Methods section, listed
#' by the type of function and the function/method name.
#'
#' @section Available Methods:
#'
#' The methods are broken into three types, based on what fields they define.
#' Different methods are used to define "bval" (the baseline value), "pval"
#' (the positive control value), and "resp" (the final response value).
#'
#' Although it does not say so specifically in each description, all methods
#' are applied by aeid.
#'
#' More information about the level 3 multiple-concentration processing is
#' available in the package vignette, "Pipeline_Overview."
#'
#' \subsection{bval Methods}{
#'   \describe{
#'     \item{bval.apid.nwlls.med}{Calculate bval as the median of cval for
#'     wells with wllt equal to "n," by apid.}
#'     \item{bval.apid.1owconc.med}{Calculate bval as the median of cval for
#'     wells with wllt equal to "t" and cndx equal to 1 or 2, by apid.}
#'     \item{bval.apid.twlls.med}{Calculate bval as the median of cval for
#'     wells with wllt equal to "t," by apid.}
#'     \item{bval.apid.tn.med}{Calculate bval as the median of cval for wells
#'     with wllt equal to "t" or "n," by apid.}
#'     \item{bval.apid.nwllslowconc.med}{Calculate bval as the median of cval
#'     for wells with wllt equal to "n" or wells with wllt equal to "t" and
#'     cndx equal to 1 or 2, by apid.}
#'     \item{bval.spid.lowconc.med}{Calculate bval as the median of cval for
#'     wells with wllt equal to "t" and cndx equal to 1, 2, or 3, by spid.}
#'   }
#' }
#'
#' \subsection{pval Methods}{
#'   \describe{
#'     \item{pval.apid.pwlls.med}{Calculate pval as the median of cval for
#'     wells with wllt equal to "p," by apid.}
#'     \item{pval.apid.mwlls.med}{Calculate pval as the median of cval for
#'     wells with wllt equal to "m," by apid.}
#'     \item{pval.apid.medpcbyconc.max}{First calculate the median of cval for
#'     wells with wllt equal to "p" or "c," by wllt, conc, and apid. Then
#'     calculate pval as the maximum of the calculated medians, by apid.}
#'     \item{pval.apid.medpcbyconc.min}{First calculate the median of cval for
#'     wells with wllt equal to "p" or "c," by wllt, conc, and apid. Then
#'     calculate pval as the minimum of the calculated medians, by apid.}
#'     \item{pval.apid.medncbyconc.min}{First calculate the median of cval for
#'     wells with wllt equal to "m" or "o," by wllt, conc, and apid. Then
#'     calculate pval as the minimum of the calculated medians, by apid.}
#'     \item{pval.apid.pmv.min}{First calculate the median of cval for
#'     wells with wllt equal to "p," "m," or "v," by wllt, conc, and apid. Then
#'     calculate pval as the minimum of the calculated medians, by apid.}
#'     \item{pval.apid.pmv.max}{First calculate the median of cval for
#'     wells with wllt equal to "p," "m," or "v," by wllt, conc, and apid. Then
#'     calculate pval as the maximum of the calculated medians, by apid.}
#'     \item{pval.apid.f.max}{First calculate the median of cval for
#'     wells with wllt equal to "f," by wllt, conc, and apid. Then calculate
#'     pval as the maximum of the calculated medians, by apid.}
#'     \item{pval.apid.f.min}{First calculate the median of cval for
#'     wells with wllt equal to "f," by wllt, conc, and apid. Then calculate
#'     pval as the minimum of the calculated medians, by apid.}
#'     \item{pval.apid.p.max}{First calculate the median of cval for
#'     wells with wllt equal to "p," by wllt, conc, and apid. Then calculate
#'     pval as the maximum of the calculated medians, by apid.}
#'     \item{pval.apid.p.min}{First calculate the median of cval for
#'     wells with wllt equal to "p," by wllt, conc, and apid. Then calculate
#'     pval as the minimum of the calculated medians, by apid.}
#'     \item{pval.apid.v.min}{First calculate the median of cval for
#'     wells with wllt equal to "v," by wllt, conc, and apid. Then calculate
#'     pval as the minimum of the calculated medians, by apid.}
#'     \item{pval.zero}{Define pval as 0.}
#'   }
#' }
#'
#' \subsection{resp Methods}{
#'   \describe{
#'     \item{resp.pc}{Calculate resp as \eqn{\frac{\mathit{cval} -
#'     \mathit{bval}}{\mathit{pval} - \mathit{bval}}100}{(cval - bval)/(pval
#'     - bval)*100}.}
#'     \item{resp.fc}{Calculate resp as \eqn{\mathit{cval}/\mathit{bval}}{
#'     cval/bval}.}
#'     \item{resp.logfc}{Calculate resp as \eqn{\mathit{cval} - \mathit{bval}}{
#'     cval - bval}.}
#'     \item{resp.log2}{Take the logarithm of resp with base 2.}
#'     \item{resp.mult25}{Multiply resp by 25.}
#'     \item{resp.scale.mad.log2fc}{Multiply resp by the scale factor
#'     \eqn{\frac{\log_2 \left( 1.2 \right)}{3\mathit{bmad}}}{
#'     log2(1.2)/(3*bmad)}.}
#'     \item{resp.scale.quant.log2fc}{Determine the maximum response
#'     \eqn{\mathit{md}}{md} where \eqn{\mathit{md}}{md} = abs(1st centile -
#'     50th centile) or abs(99th centile - 50th centile), whichever is greater.
#'     Scale the response such that 20 percent of md equals
#'     \eqn{\log_2 \left( 1.2 \right)}{log2(1.2)}.}
#'     \item{resp.multneg1}{Multiply resp by -1.}
#'     \item{resp.shiftneg.3bmad}{Shift all resp values less than 3*bmad to 0.}
#'     \item{resp.shiftneg.6bmad}{Shift all resp values less than 6*bmad to 0.}
#'     \item{resp.shiftneg.10bmad}{Shift all resp values less than 10*bmad to
#'     0.}
#'     \item{resp.blineshift.3bmad.repi}{Shift resp values with the blineShift
#'     function by repi, where the window (wndw) is 3*bmad.}
#'     \item{resp.blineshift.50.repi}{Shift resp values with the blineShift
#'     function by repi, where the window (wndw) is 50.}
#'     \item{resp.blineshift.3bmad.spid}{Shift resp values with the blineShift
#'     function by spid, where the window (wndw) is 3*bmad.}
#'     \item{resp.blineshift.50.spid}{Shift resp values with the blineShift
#'     function by spid, where the window (wndw) is 50.}
#'     \item{none}{Do no normalization; make resp equal to cval.}
#'   }
#' }
#'
#' @keywords internal
#' 
#' @note
#' This function is not exported and is not intended to be used by the user.


mc3_mthds <- function() {

    list(

        bval.apid.nwlls.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(cval[wllt == "n"], na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        bval.apid.1owconc.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(
                        cval[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        bval.apid.twlls.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(cval[wllt == "t"], na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        bval.apid.tn.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(
                        cval[wllt %in% c("t", "n")],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        bval.apid.nwllslowconc.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(
                        cval[(cndx %in% seq_len(2) & 
                            wllt == "t") |
                            wllt == "n"],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        bval.spid.lowconc.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(
                        cval[cndx %in% seq_len(3) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ),
                    by=list(aeid, spid, repi)]
            )
            list(e1)

        },

        pval.apid.pwlls.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    pval := median(cval[wllt == "p"], na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        pval.apid.mwlls.med=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    pval := median(cval[wllt == "m"], na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            list(e1)

        },

        pval.apid.medpcbyconc.max=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(
                        cval[wllt %in% c("c", "p")],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid, wllt, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := max(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.medpcbyconc.min=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(
                        cval[wllt %in% c("c", "p")],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid, wllt, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := min(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.medncbyconc.min=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(
                        cval[wllt %in% c("m","o")],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid, wllt, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := min(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.pmv.min=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(
                        cval[wllt %in% c("p", "m", "v")],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid, wllt, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := min(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.pmv.max=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(
                        cval[wllt %in% c("p", "m", "v")],
                        na.rm=TRUE
                    ),
                    by=list(aeid, apid, wllt, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := max(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.f.max=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(cval[wllt == "f"], na.rm=TRUE),
                    by=list(aeid, apid, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := max(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.f.min=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(cval[wllt == "f"], na.rm=TRUE),
                    by=list(aeid, apid, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := min(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.p.max=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(cval[wllt == "p"], na.rm=TRUE),
                    by=list(aeid, apid, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := max(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.p.min=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(cval[wllt == "p"], na.rm=TRUE),
                    by=list(aeid, apid, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := min(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.apid.v.min=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    temp := median(cval[wllt == "v"], na.rm=TRUE),
                    by=list(aeid, apid, conc)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    pval := min(temp, na.rm=TRUE),
                    by=list(aeid, apid)]
            )
            e3 <- bquote(dat[ , temp := NULL])
            list(e1, e2, e3)

        },

        pval.zero=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), pval := 0])
            list(e1)

        },

        resp.pc=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    resp := (cval - bval)/(pval - bval)*100]
            )
            list(e1)

        },

        resp.fc=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), resp := cval/bval])
            list(e1)

        },

        resp.logfc=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), resp := cval - bval])
            list(e1)

        },

        resp.log2=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), resp := log2(resp)])
            list(e1)

        },

        resp.mult25=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), resp := resp * 25])
            list(e1)

        },

        resp.scale.mad.log2fc=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bmad := mad(
                        resp[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ),
                    by=aeid]
            )
            e2 <- bquote(dat[J(.(aeids)), resp := log2(1.2)/(3*bmad)*resp])
            e3 <- bquote(dat[ , bmad := NULL])

            list(e1, e2, e3)

        },

        resp.scale.quant.log2fc=function(aeids) {

            qv <- c(0.01, 0.5, 0.99)
            e1 <- bquote(
                dat[J(.(aeids)),
                    c("q1", "q2", "q3") := as.list(
                        quantile(resp, .(qv))),
                    by=aeid]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    md := max(
                        abs(
                            c(
                                diff(c(q1, q2)),
                                diff(c(q2, q3))
                            )
                        )
                    ),
                    by=aeid]
            )
            e3 <- bquote(dat[J(.(aeids)), resp := log2(1.2)/(0.2*md)*resp])
            e4 <- bquote(dat[ , .(c("q1", "q2", "q3", "md")) := NULL])
            list(e1, e2, e3, e4)

        },

        resp.multneg1=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), resp := resp * -1])
            list(e1)

        },

        resp.shiftneg.3bmad=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bmad := mad(
                        resp[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ),
                    by=aeid]
            )
            e2 <- bquote(dat[aeid %in% .(aeids) & resp < -3 * bmad, resp := 0])
            e3 <- bquote(dat[ , bmad := NULL])
            list(e1, e2, e3)

        },

        resp.shiftneg.6bmad=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bmad := mad(
                        resp[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ),
                    by=aeid]
            )
            e2 <- bquote(dat[aeid %in% .(aeids) & resp < -6 * bmad, resp := 0])
            e3 <- bquote(dat[ , bmad := NULL])
            list(e1, e2, e3)

        },

        resp.shiftneg.10bmad=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bmad := mad(
                        resp[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ),
                    by=aeid]
            )
            e2 <- bquote(dat[aeid %in% .(aeids) & resp < -10 * bmad, resp := 0])
            e3 <- bquote(dat[ , bmad := NULL])
            list(e1, e2, e3)

        },

        resp.blineshift.3bmad.repi=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    wndw := mad(
                        resp[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ) * 3,
                    by=aeid]
            )
            e2 <- bquote(
                dat[aeid %in% .(aeids) & wllt %in% c("t", "c", "o"),
                    resp := blineShift(resp, logc, wndw),
                    by=list(aeid, spid, repi)]
            )
            e3 <- bquote(dat[ , wndw := NULL])
            list(e1, e2, e3)

        },

        resp.blineshift.50.repi=function(aeids) {

            e1 <- bquote(
                dat[aeid %in% .(aeids) & wllt %in% c("t", "c", "o"),
                    resp := blineShift(resp, logc, wndw=50),
                    by=list(aeid, spid, repi)]
            )
            list(e1)

        },

        resp.blineshift.3bmad.spid=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    wndw := mad(
                        resp[cndx %in% seq_len(2) & 
                            wllt == "t"],
                        na.rm=TRUE
                    ) * 3,
                    by=aeid]
            )
            e2 <- bquote(
                dat[aeid %in% .(aeids) & wllt %in% c("t", "c", "o"),
                    resp := blineShift(resp, logc, wndw),
                    by=list(aeid, spid)]
            )
            e3 <- bquote(dat[ , wndw := NULL])
            list(e1, e2, e3)

        },

        resp.blineshift.50.spid=function(aeids) {

            e1 <- bquote(
                dat[aeid %in% .(aeids) & wllt %in% c("t", "c", "o"),
                    resp := blineShift(resp, logc, wndw=50),
                    by=list(aeid, spid)]
            )
            list(e1)

        },

        none=function(aeids) {

            e1 <- bquote(dat[J(.(aeids)), resp := cval])
            list(e1)

        },

        ## PMI-specific methods

        bval.pmi=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    bval := median(cval[wllt == "n"], na.rm=TRUE),
                    by=list(aeid, apid, vhid)]
            )
            list(e1)

        },

        resp.pmi=function(aeids) {

            e1 <- bquote(
                dat[J(.(aeids)),
                    tmp := unique(bval[wllt == "c"]),
                    by=list(apid)]
            )
            e2 <- bquote(
                dat[J(.(aeids)),
                    resp := (cval - bval)/(pval - tmp)*100]
            )
            e3 <- bquote(dat[ , tmp := NULL])
            list(e1, e2, e3)

        }

    )

}

#-------------------------------------------------------------------------------