Nothing
R/Statistics.R
In dowser: B Cell Receptor Phylogenetics Toolkit
Defines functions
calcRF
correlationTest
runCorrelationTest
resolvePolytomies
getDivergence
testSC
testSP
testPS
Documented in
calcRF
correlationTest
getDivergence
resolvePolytomies
runCorrelationTest
testPS
testSC
testSP
# Summary statistics and data transformation for switch count distributions
#' Performs PS (parsimony score) test on switch data
#'
#' \code{testPS} performs a PS test
#' @param switches Data frame from findSwitches
#' @param bylineage Perform test for each lineage individually? (FALSE)
#' @param pseudocount Pseudocount for P value calculations
#' @param alternative Perform one-sided (\code{greater} or \code{less})
#' or \code{two.sided} test
#'
#' @return A list containing a \code{tibble} with mean PS statistics, and another
#' with PS statistics per repetition.
#' @details
#' Output data table columns:
#' RECON = PS for observed data
#' PERMUTE = PS for permuted data
#' DELTA = RECON - PERMUTE
#' PLT = p value for DELTA < 0
#' PGT = p value for DELTA < 0
#' \itemize{
#' \item \code{RECON}: PS for observed data.
#' \item \code{PERMUTE}: PS for permuted data.
#' \item \code{DELTA}: RECON - PERMUTE.
#' \item \code{PLT}: p value that DELTA < 0
#' \item \code{PGT}: p value that DELTA > 0
#' \item \code{STAT}: Statistic used (PS).
#' \item \code{REP}: Bootstrap repetition.
#' \item \code{REPS}: Total number of ootstrap repetition.
#'}
#'
#' @seealso Uses output from \link{findSwitches}. Related to \link{testSP}
#' and \link{testSC}.
#' @examples
#' \dontrun{
#' igphyml <- "~/apps/igphyml/src/igphyml"
#' data(ExampleAirr)
#' ExampleAirr$sample_id <- sample(ExampleAirr$sample_id)
#' clones <- formatClones(ExampleAirr, trait="sample_id")
#' btrees <- findSwitches(clones[1:2], bootstraps=10, nproc=1,
#' igphyml=igphyml, trait="sample_id")
#' testPS(btrees$switches)
#' }
#' @export
testPS <- function(switches, bylineage=FALSE, pseudocount=0,
alternative=c("less","two.sided","greater")){
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N" &
!!rlang::sym("TO") != !!rlang::sym("FROM") &
!!rlang::sym("FROM") != "UCA")
if(nrow(switches) == 0){
stop("No switches left after filtering!")
}
if(!bylineage){
reps <- switches %>%
dplyr::group_by(!!rlang::sym("REP"), !!rlang::sym("TYPE")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::ungroup() %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("SWITCHES")) %>%
dplyr::mutate(DELTA = !!rlang::sym("RECON") - !!rlang::sym("PERMUTE"))
}else{
reps <- switches %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("CLONE"),
!!rlang::sym("REP")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::ungroup() %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("SWITCHES")) %>%
dplyr::mutate(DELTA = !!rlang::sym("RECON") - !!rlang::sym("PERMUTE"))
}
if(!bylineage){
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}else{
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}
means$STAT <- "PS"
reps$STAT <- "PS"
means$REPS <- length(unique(reps$REP))
results <- list()
results$reps <- reps
results$means <- means
results
}
#' Performs SP (switch proportion) test on switch data
#'
#' \code{testSP} performs an SP test
#' @param switches Data frame from findSwitches
#' @param permuteAll Permute among trees?
#' @param from Include only switches from this state?
#' @param to Include only switches to this state?
#' @param dropzeroes Drop switches with zero counts?
#' @param bylineage Perform test for each lineage individually?
#' @param pseudocount Pseudocount for P value calculations
#' @param alternative Perform one-sided (\code{greater} or \code{less})
#' or \code{two.sided} test
#' @param tip_switch maximum tip/switch ratio
#' @param exclude exclude clones with tip/switch ratio > \code{tip_switch}?
#' @return A list containing a \code{tibble} with mean SP statistics, and another
#' with SP statistics per repetition.
#'
#' @details
#' Output data table columns:
#' RECON = SP for observed data
#' PERMUTE = SP for permuted data
#' DELTA = RECON - PERMUTE
#' PLT = p value for DELTA < 0
#' PGT = p value for DELTA < 0
#' \itemize{
#' \item \code{FROM}: State going from.
#' \item \code{TO}: State going to.
#' \item \code{RECON}: SP for observed data.
#' \item \code{PERMUTE}: SP for permuted data.
#' \item \code{DELTA}: RECON - PERMUTE.
#' \item \code{PLT}: p value that DELTA < 0
#' \item \code{PGT}: p value that DELTA > 0
#' \item \code{STAT}: Statistic used (SP).
#' \item \code{REP}: Bootstrap repetition.
#' \item \code{REPS}: Total number of ootstrap repetition.
#'}
#'
#' @seealso Uses output from \link{findSwitches}. Related to \link{testPS}
#' and \link{testSC}.
#' @examples
#' \dontrun{
#' igphyml <- "~/apps/igphyml/src/igphyml"
#' data(ExampleAirr)
#' ExampleAirr$sample_id = sample(ExampleAirr$sample_id)
#' clones = formatClones(ExampleAirr, trait="sample_id")
#' btrees = findSwitches(clones[1:2], bootstraps=10, nproc=1,
#' igphyml=igphyml, trait="sample_id")
#' testSP(btrees$switches)
#' }
#' @export
testSP <- function(switches, permuteAll=FALSE,
from=NULL, to=NULL, dropzeroes=TRUE,
bylineage=FALSE, pseudocount=0, alternative=c("greater","two.sided","less"),
tip_switch=20, exclude=FALSE){
permute <- dplyr::quo(!!rlang::sym("PERMUTE"))
if(permuteAll){
permute <- dplyr::quo(!!rlang::sym("PERMUTEALL"))
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != !!rlang::sym("FROM") &
!!rlang::sym("FROM") != "UCA")
if(!is.null(from)){
from <- dplyr::enquo(from)
switches <- dplyr::filter(switches, !!rlang::sym("FROM") == !!from)
}
if(!is.null(to)){
to <- dplyr::enquo(to)
switches <- dplyr::filter(switches, !!rlang::sym("TO") == !!to)
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N")
if(nrow(switches) == 0){
stop("No switches left after filtering!")
}
if(exclude){
# separate tips
tips <- dplyr::filter(switches, !!rlang::sym("TO")=="NTIP" &
!!rlang::sym("TYPE")=="RECON")
# germline doesn't count in downsampling algorithm
tips$SWITCHES <- tips$SWITCHES - 1
counts <- testPS(switches, bylineage=TRUE)$means
m <- match(counts$CLONE, tips$CLONE)
counts$tips <- tips[m,]$SWITCHES
counts$ratio <- counts$tips/counts$RECON
excluded <- dplyr::filter(counts, !!rlang::sym("ratio") > tip_switch)$CLONE
if(length(excluded) > 0){
warning(paste("Excluding clone(s)",paste(excluded,collapse=",")
,"due to high tip/switch ratio"))
switches <- dplyr::filter(switches,!(!!rlang::sym("CLONE") %in% excluded))
}
}
if(!bylineage){
reps <- switches %>%
dplyr::group_by(!!rlang::sym("REP"), !!rlang::sym("TYPE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES")))
reps <- reps %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("REP")) %>%
dplyr::mutate(PROP = !!rlang::sym("SWITCHES")/
sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("PROP"))
}else{
reps <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("CLONE"),
!!rlang::sym("REP")) %>%
dplyr::mutate(PROP = !!rlang::sym("SWITCHES")/
sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("PROP"))
}
reps <- reps %>% dplyr::mutate(DELTA = !!rlang::sym("RECON")-!!permute)
if(dropzeroes){
from_type <-
reps %>% dplyr::group_by(!!rlang::sym("FROM")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("FROM"))
to_type <-
reps %>% dplyr::group_by(!!rlang::sym("TO")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("TO"))
remove <- to_type[to_type %in% from_type]
reps <- reps[!reps$FROM %in% remove &
!reps$TO %in% remove, ]
}
if(sum(is.na(reps$DELTA)) > 0){
na_reps <- unique(reps[is.na(reps$DELTA),]$REP)
warning(paste0("NA delta values in ",
length(na_reps)," replicates, discarding"))
reps <- dplyr::filter(reps, !(!!rlang::sym("REP") %in% na_reps))
}
if(!bylineage){
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}else{
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}
means$STAT <- "SP"
reps$STAT <- "SP"
means$REPS <- length(unique(reps$REP))
results <- list()
results$reps <- reps
results$means <- means
results
}
#' Performs SC (switch count) test on switch data
#'
#' \code{testSC} performs an SC test
#' @param switches Data frame from findSwitches
#' @param permuteAll Permute among trees?
#' @param from Include only switches from this state?
#' @param to Include only switches to this state?
#' @param dropzeroes Drop switches with zero counts?
#' @param bylineage Perform test for each lineage individually?
#' @param pseudocount Pseudocount for P value calculations
#' @param alternative Perform one-sided (\code{greater} or \code{less})
#' or \code{two.sided} test
#' @return A list containing a \code{tibble} with mean SC statistics, and another
#' with SC statistics per repetition.
#'
#' @details
#' Output data table columns:
#' RECON = SC for observed data
#' PERMUTE = SC for permuted data
#' DELTA = RECON - PERMUTE
#' PLT = p value for DELTA < 0
#' PGT = p value for DELTA < 0
#' \itemize{
#' \item \code{FROM}: State going from.
#' \item \code{TO}: State going to.
#' \item \code{RECON}: SC for observed data.
#' \item \code{PERMUTE}: SC for permuted data.
#' \item \code{DELTA}: RECON - PERMUTE.
#' \item \code{PLT}: p value that DELTA < 0
#' \item \code{PGT}: p value that DELTA > 0
#' \item \code{STAT}: Statistic used (SC).
#' \item \code{REP}: Bootstrap repetition.
#' \item \code{REPS}: Total number of ootstrap repetition.
#'}
#'
#' @seealso Uses output from \link{findSwitches}. Related to \link{testPS}
#' and \link{testSP}.
#' @examples
#' \dontrun{
#' igphyml <- "~/apps/igphyml/src/igphyml"
#' data(ExampleAirr)
#' ExampleAirr$sample_id = sample(ExampleAirr$sample_id)
#' clones = formatClones(ExampleAirr, trait="sample_id")
#' btrees = findSwitches(clones[1:2], bootstraps=100, nproc=1,
#' igphyml=igphyml, trait="sample_id", id="temp", dir="temp")
#' testSC(btrees$switches)
#' }
#' @export
testSC <- function(switches,dropzeroes=TRUE,
bylineage=FALSE, pseudocount=0, from=NULL, to=NULL,
permuteAll=FALSE, alternative=c("two.sided","greater","less")){
permute <-dplyr::quo(!!rlang::sym("PERMUTE"))
if(permuteAll){
permute <-dplyr::quo(!!rlang::sym("PERMUTEALL"))
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != !!rlang::sym("FROM") &
!!rlang::sym("FROM") != "UCA")
if(!is.null(from)){
from <-dplyr::enquo(from)
switches <- dplyr::filter(switches, !!rlang::sym("FROM") == !!from)
}
if(!is.null(to)){
to <-dplyr::enquo(to)
switches <- dplyr::filter(switches, !!rlang::sym("TO") == !!to)
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N")
if(nrow(switches) == 0){
stop("No switches left after filtering!")
}
if(!bylineage){
reps <- switches %>%
dplyr::group_by(!!rlang::sym("REP"), !!rlang::sym("TYPE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES")))
reps <- reps %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("REP")) %>%
dplyr::mutate(COUNT = !!rlang::sym("SWITCHES")) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("COUNT"))
}else{
reps <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("CLONE"),
!!rlang::sym("REP")) %>%
dplyr::mutate(COUNT = !!rlang::sym("SWITCHES")) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("COUNT"))
}
reps <- reps %>% dplyr::mutate(DELTA = !!rlang::sym("RECON")-!!permute)
if(dropzeroes){
from_type <-
reps %>% dplyr::group_by(!!rlang::sym("FROM")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("FROM"))
to_type <-
reps %>% dplyr::group_by(!!rlang::sym("TO")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("TO"))
remove <- to_type[to_type %in% from_type]
reps <- reps[!reps$FROM %in% remove &
!reps$TO %in% remove, ]
}
if(!bylineage){
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}else{
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}
means$STAT <- "SC"
reps$STAT <- "SC"
means$REPS <- length(unique(reps$REP))
results <- list()
results$reps <- reps
results$means <- means
results
}
#' Get divergence from root of tree for each tip
#'
#' \code{getDivergence} get sum of branch lengths leading from the
#' root of the tree. If the germline sequence is included in the tree,
#' this will equal the germline divergence. If germline removed,
#' this will equal the MRCA divergence
#' @param phy Tree object
#' @param minlength Branch lengths to collapse in trees
#' @return A named vector of each tip's divergence from the tree's root.
#'
#' @export
getDivergence = function(phy, minlength=0.001){
tips <- phy$tip.label
phy$edge.length[phy$edge.length < minlength] <- 0
phy <- ape::di2multi(phy,tol=minlength)
uca <- ape::getMRCA(phy,tip=tips)
co <- ape::dist.nodes(phy)
dist <- co[1:length(tips),uca]
names(dist) <- tips
dist
}
#' Resolve polytomies to have the minimum number of single timepoint clades
#'
#' @param phy Tree object
#' @param clone airrClone data object corresponding to \code{phy}
#' @param time Column name holding numeric time information
#' @param sequence Column name holding sequence ID
#' @param germline Germline sequence name
#' @param minlength Branch lengths to collapse in trees
#' @param verbose Print lots of rubbish while running?
#' @return A \code{phylo} tree object in which polytomies are resolved to
#' have the minimum number of single timepoint clades.
#'
#' @details
#' Iteratively identifies polytomies (clusters of < minlength branches),
#' prunes each descendant branch, combines clades with the same timepoint
#' before grouping them back together. Checks to make sure that the divergence
#' of each tip is the same after resolution.
#'
#' @seealso Uses output from \link{getTrees} during \link{correlationTest}.
#' @export
resolvePolytomies = function(phy, clone, minlength=0.001,
time="time", sequence="sequence_id", germline = "Germline",
verbose=FALSE){
data <- clone@data
phy <- rerootTree(ape::di2multi(phy,tol=minlength),
germline=germline,verbose=0)
tips <- phy$tip.label
if(sum(!data[[sequence]] %in% phy$tip.label) > 0){
stop("Tree and data sequence ids don't match")
}
ittybitty_branch = 1e-7
# ape does weird things with zero branches
phy$edge.length[phy$edge.length == 0] = ittybitty_branch
odivergence <- getDivergence(phy,minlength)
polytomies <- table(phy$edge[,1])
polytomies <- names(polytomies[polytomies > 2])
while(length(polytomies) > 0){
if(verbose){
print(polytomies)
}
# for first polytomy in list, remove each subclade
# and record the time of its sequences
target_node <- as.numeric(polytomies[1])
node_index <- which(phy$edge[,1] == target_node)
node_ns <- phy$edge[node_index,2]
edge_ls <- phy$edge.length[node_index]
clades <- list()
times <- list()
for(i in 1:length(node_ns)){
nc <- node_ns[i]
# unfortuantely processing is different if clade is > 1 tip.
if(nc <= length(phy$tip.label)){ #clade is a tip
clades[[i]] <- ape::keep.tip(phy, tip = nc)
times[[i]] <- unique(data[data[[sequence]] %in%
clades[[i]]$tip.label,][[time]])
clades[[i]]$root.edge <- edge_ls[[i]] #store root edge
clades[[i]]$edge.length <- 0
}else{
tips <- ape::extract.clade(phy, node = nc,
collapse.singles=TRUE)$tip.label
clades[[i]] <- ape::keep.tip(phy, tip = tips)
clades[[i]]$root.edge <- edge_ls[[i]] #store root edge
times[[i]] <- unique(data[data[[sequence]] %in%
clades[[i]]$tip.label,][[time]])
if(length(times[[i]]) > 1){
times[[i]] <- "multi"
}
}
}
# combine all clades of each time type in a balanced
# manner, so they remain clades
topclades <- list()
distinct_times <- unique(unlist(times))
for(ctime in distinct_times){
tclades <- clades[unlist(lapply(times,
function(x)x==ctime))]
clade <- tclades[[1]]
if(length(tclades) > 1){
for(i in 2:length(tclades)){
clade <- ape::bind.tree(clade,tclades[[i]],
position=clade$root.edge)
clade <- ape::collapse.singles(clade)
clade$root.edge <- ittybitty_branch
}
}
topclades[[as.character(ctime)]] <- clade
}
# combine top clades into a polytomy
polytomy <- topclades[[1]]
if(length(topclades) > 1){
for(i in 2:length(topclades)){
polytomy <- ape::bind.tree(polytomy,topclades[[i]],
position=polytomy$root.edge)
polytomy <- ape::collapse.singles(polytomy)
polytomy$root.edge <- ittybitty_branch
}
}
tphy <- ape::drop.tip(phy, collapse.singles=TRUE, trim.internal=FALSE, tip =
ape::extract.clade(phy, node = target_node)$tip.label, subtree=TRUE)
pruned_tip <- paste0("[",length(polytomy$tip.label),"_tips]")
ntree <- ape::bind.tree(tphy, polytomy, where=which(tphy$tip.label==pruned_tip),
position=polytomy$root.edge)
ntree <- ape::drop.tip(ntree, tip=pruned_tip)
phy <- ntree
polytomies <- table(phy$edge[,1])
polytomies <- names(polytomies[polytomies > 2])
}
ndivergence <- getDivergence(phy,minlength)
if(length(ndivergence) != length(odivergence)){
stop("Number of tips corrupted during polytomy resolution")
}
if(sum(odivergence[names(ndivergence)] != ndivergence)){
stop("Divergence corrupted during polytomy resolution")
}
phy
}
#' Run correlationTest, based on https://doi.org/10.1111/2041-210X.12466
#'
#' \code{runCorrelationTest} performs root-to-tip regression permutation test
#' @param phy Tree object
#' @param clone airrClone data object corresponding to \code{phy}
#' @param permutations Number of permutations to run
#' @param polyresolve Resolve polytomies to have a minimum number of
#' single timepoint clades
#' @param permutation Permute among single timepoint clades or uniformly
#' among tips
#' @param time Column name holding numeric time information
#' @param sequence Column name holding sequence ID
#' @param germline Germline sequence name
#' @param minlength Branch lengths to collapse in trees
#' @param verbose Print lots of rubbish while running?
#' @param alternative Is alternative that the randomized correlation are greater than
#' or equal to observed, or greater/less than?
#' @return A list of statistics from running the permutation test.
#'
#' @details
#' See \link{correlationTest} for details
#' @seealso \link{correlationTest}.
runCorrelationTest = function(phy, clone, permutations, minlength=0.001,
polyresolve = TRUE, permutation = c("clustered", "uniform"),
time="time", sequence="sequence_id", germline = "Germline",
verbose=TRUE, alternative = c("greater","two.sided")){
data <- clone@data
if(verbose){
print(paste("Analyzing clone: ",clone@clone))
}
if(polyresolve){
if(verbose){
print("resolving polytomies to single timepoint clades")
}
phy <- resolvePolytomies(phy, clone, minlength,
time=time, sequence=sequence, germline = germline,
verbose=verbose)
}
phy <- ape::drop.tip(phy,germline)
tips <- phy$tip.label
if(sum(!data[[sequence]] %in% phy$tip.label) > 0){
stop("Tree and data sequence ids don't match")
}
# dates is named and ordered by tree tips
dates <- data[[time]]
names(dates) <- data[[sequence]]
dates <- dates[tips]
if(dplyr::n_distinct(dates)==1){
warning(paste("Only one timepoint cluster in clone",clone@clone))
results <- list(correlation=NA)
results[["clone"]] <- clone
results[["tree"]] <- phy
results[["random"]] <- NA
results[["random_correlation"]] <- NA
results[["p_gt"]] <- NA
results[["p_lt"]] <- NA
results[["nposs"]] <- NA
results[["nclust"]] <- NA
results[["p"]] <- NA
results[["min_p"]] <- NA
results[["slope"]] <- NA
return(results)
}
divergence <- getDivergence(phy, minlength)
data$divergence <- divergence[data[[sequence]]]
cl <- 1:length(dates) # each tip is its own cluster
if(permutation[1] == "clustered"){
# define all monophyletic clades with > 1 tips
gc <- lapply(ape::subtrees(phy),function(x)x$tip.label)
# identify which clades are single timepoint
single.date.clades <- which(unlist(lapply(gc,function(x)
dplyr::n_distinct(dates[x])))==1)
if(length(single.date.clades)>0){
for(i in 1:length(single.date.clades)){
m <- match(gc[[single.date.clades[i]]],phy$tip.label)
cl[m] <- cl[m[1]] # assign clusters to be the same across single timepoint clades
}
}
cl <- match(cl,unique(cl)) #make clusters 1:(number of clusters)
}
# assign clusters to each sequence
names(cl) <- tips
m <- match(data[[sequence]],names(cl))
data$cluster <- cl[data$sequence_id]
counts <- table(data$cluster,data[[time]])
if(sum(rowSums(counts > 0) > 1) > 0){
stop("Clusters are not single timepoint!")
}
# get ordered list of times for each cluster
ctimes <- unlist(lapply(sort(unique(data$cluster)), function(x)
unique(data[data$cluster == x,][[time]])))
true <- stats::cor(data[[time]],data$divergence)
random <- rep(0, length=permutations)
for(i in 1:permutations){
times <- ctimes[sample(1:length(ctimes))]
data$random_time <- times[data$cluster]
random[i] <- stats::cor(data$divergence, data$random_time)
}
nclust <- dplyr::n_distinct(cl)
nposs <-exp(lfactorial(nclust)-sum(lfactorial(table(ctimes))))
minp <- 1
if(alternative[1] == "two.sided"){
minp <- 0.5
}
clone@data <- data
slope <- summary(stats::lm(data$divergence ~ data[[time]]))$coefficients[2,1]
results <- list(correlation=true)
results[["clone"]] <- clone
results[["tree"]] <- phy
results[["random"]] <- random
results[["random_correlation"]] <- mean(random)
random <- c(random,true)
results[["p_gt"]] <- (sum(true < random) +
sum(true == random)*0.5)/length(random)
results[["p_lt"]] <- (sum(true > random) +
sum(true == random)*0.5)/length(random)
results[["nposs"]] <- nposs
results[["nclust"]] <- nclust
results[["p"]] <- mean(true <= random)
results[["min_p"]] <- max(minp/nposs, minp/(permutations+1))
results[["slope"]] <- slope
results
}
#' Run date randomization test for temporal signal on a set of trees.
#'
#' \code{correlationTest} performs root-to-tip regression date randomization test
#' @param clones A \code{tibble} object containing airrClone and \code{phylo} objects
#' @param permutations Number of permutations to run
#' @param polyresolve Resolve polytomies to have a minimum number of
#' single timepoint clades
#' @param perm_type Permute among single timepoint clades or uniformly
#' among tips
#' @param time Column name holding numeric time information
#' @param sequence Column name holding sequence ID
#' @param germline Germline sequence name
#' @param minlength Branch lengths to collapse in trees
#' @param verbose Print lots of rubbish while running?
#' @param storeTree Store the tree used?
#' @param alternative Is alternative that the randomized correlation are greater than
#' or equal to observed, or greater/less than?
#' @param nproc Number of cores to use for calculations. Parallelizes by tree.
#' @return A \code{tibble} with the same columns as clones, but additional
#' columns corresponding to test statistics for each clone.
#'
#' @details
#' Object returned contains these columns which are added or modified from input:
#' \itemize{
#' \item \code{data}: airrClone object, same as input but with additional columns
#' "cluster" which correspond to permutation cluster, and "divergence."
#' \item \code{slope}: Slope of linear regression between divergence and time.
#' \item \code{correlation}: Correlation between divergence and time.
#' \item \code{p}: p value of correlation compared to permuted correlations.
#' \item \code{random_correlation}: Mean correlation of permutation replicates.
#' \item \code{min_p}: Minimum p value of data, determined by either the number of
#' distinct clade/timepoint combinations or number of permutations.
#' \item \code{nposs}: Number of possible distinct timepoint/clade combinations.
#' \item \code{nclust}: Number of clusters used in permutation. If perm_type="uniform"
#' this is the number of tips.
#' \item \code{p_gt/p_lt}: P value that permuted correlations are greater or less
#' than observed correlation. Only returned if alternative = "two.sided"
#' \item \code{test_trees}: The \link{phylo} tree objects used, possibly with
#' resolved polytomies.
#' }
#' @seealso Uses output from \code{getTrees}.
#' @export
correlationTest = function(clones, permutations=1000, minlength=0.001,
perm_type = c("clustered", "uniform"), time="time",
sequence="sequence_id", germline = "Germline",
verbose=FALSE, polyresolve = TRUE,
alternative = c("greater","two.sided"),
storeTree = FALSE, nproc=1){
if(!"tbl" %in% class(clones)){
print(paste("clones is of class",class(clones)))
stop("clones must be a tibble of airrClone objects!")
}else{
if(!inherits(clones$data[[1]], "airrClone")){
print(paste("clones is list of class",class(clones$data[[1]])))
stop("clones must be a list of airrClone objects!")
}
}
if(!"trees" %in% names(clones)){
stop("clones must have trees column!")
}
time_check <- unlist(lapply(clones$data, function(x)!time %in% names(x@data)))
if(sum(time_check) > 0){
stop(paste("Time column",time,"not found in clone object (must be trait value in formatClones)"))
}
time_check <- unlist(lapply(clones$data, function(x)!is.numeric(x@data[[time]])))
if(sum(time_check) > 0){
stop(paste("Time column",time,"contains non-numeric values, impossible to continue"))
}
results <- parallel::mclapply(1:nrow(clones),function(x)
runCorrelationTest(clones$trees[[x]], clones$data[[x]],
permutations, minlength=minlength, polyresolve=polyresolve,
permutation=perm_type, time= time,
sequence=sequence, germline=germline,
verbose=verbose, alternative=alternative),
mc.cores=nproc)
clones$data <- lapply(results,function(x)x$clone)
if(storeTree){
clones$test_trees <- lapply(results,function(x)x$tree)
}
cols <- c("slope", "p", "correlation", "random_correlation",
"min_p", "nposs", "nclust")
if(alternative[1] == "two.sided"){
cols <- c(cols, c("p_gt", "p_lt"))
}
for(col in cols){
clones[[col]] <- unlist(lapply(results,function(x)x[[col]]))
}
clones
}
#' Finds the Robinson-Fould's cluster distance between phylogenies.
#'
#' \code{calcRF} Calculates the RF distance between two phylogenetic trees with
#' the same tips and tip labels.
#' @param tree_1 A \code{phylo} object
#' @param tree_2 A \code{phylo} object
#' @param nproc Number of cores to use for calculations.
#'
#' @return The RF cluster value for the two input trees.
#'
#' @export
calcRF <- function(tree_1, tree_2, nproc = 1){
tip_amount_check <- length(tree_1$tip.label) == length(tree_2$tip.label)
if(!tip_amount_check){
stop("trees do not have the same number of tips")
}
tip_check <- dplyr::setdiff(tree_1$tip.label, tree_2$tip.label)
# change this
if(!identical(tip_check, character(0))){
stop("tree tip labels are not identical")
}
tree_1_df <- splits_func(list(tree_1),1)
tree_2_df <- splits_func(list(tree_2), 1)
total_mismatches <- unlist(parallel::mclapply(1:nrow(tree_1_df), function(x){
tree_1_sub <- tree_1_df$found[[x]]
mismatch_vector <- unlist(lapply(1:nrow(tree_2_df), function(y){
mismatches_1 <- dplyr::setdiff(tree_2_df$found[[y]], tree_1_sub)
mismatches_2 <- dplyr::setdiff(tree_1_sub, tree_2_df$found[[y]])
if(identical(mismatches_1, character(0)) & identical(mismatches_2, character(0))){
value <- "match"
} else{
value <- "mismatch"
}
return(value)
}))
if("match" %in% mismatch_vector){
tobind <- 0
} else{
tobind <- 1
}
return(tobind)
}, mc.cores = nproc))
total_mismatches <- sum(total_mismatches)
clone_mismatches <- unlist(parallel::mclapply(1:nrow(tree_2_df), function(x){
tree_2_sub <- tree_2_df$found[[x]]
mismatch_vector <- unlist(lapply(1:nrow(tree_1_df), function(y){
mismatches_1 <- setdiff(tree_1_df$found[[y]], tree_2_sub)
mismatches_2 <- setdiff(tree_2_sub, tree_1_df$found[[y]])
if(identical(mismatches_1, character(0)) & identical(mismatches_2, character(0))){
value <- "match"
} else{
value <- "mismatch"
}
return(value)
}))
if("match" %in% mismatch_vector){
tobind <- 0
} else{
tobind <- 1
}
return(tobind)
}, mc.cores = nproc))
clone_mismatches <- sum(clone_mismatches)
all_mismatches <- total_mismatches + clone_mismatches
return(all_mismatches)
}
Try the dowser package in your browser
Any scripts or data that you put into this service are public.
dowser documentation built on May 29, 2024, 3:06 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions calcRF correlationTest runCorrelationTest resolvePolytomies getDivergence testSC testSP testPS
Documented in calcRF correlationTest getDivergence resolvePolytomies runCorrelationTest testPS testSC testSP
# Summary statistics and data transformation for switch count distributions
#' Performs PS (parsimony score) test on switch data
#'
#' \code{testPS} performs a PS test
#' @param switches Data frame from findSwitches
#' @param bylineage Perform test for each lineage individually? (FALSE)
#' @param pseudocount Pseudocount for P value calculations
#' @param alternative Perform one-sided (\code{greater} or \code{less})
#' or \code{two.sided} test
#'
#' @return A list containing a \code{tibble} with mean PS statistics, and another
#' with PS statistics per repetition.
#' @details
#' Output data table columns:
#' RECON = PS for observed data
#' PERMUTE = PS for permuted data
#' DELTA = RECON - PERMUTE
#' PLT = p value for DELTA < 0
#' PGT = p value for DELTA < 0
#' \itemize{
#' \item \code{RECON}: PS for observed data.
#' \item \code{PERMUTE}: PS for permuted data.
#' \item \code{DELTA}: RECON - PERMUTE.
#' \item \code{PLT}: p value that DELTA < 0
#' \item \code{PGT}: p value that DELTA > 0
#' \item \code{STAT}: Statistic used (PS).
#' \item \code{REP}: Bootstrap repetition.
#' \item \code{REPS}: Total number of ootstrap repetition.
#'}
#'
#' @seealso Uses output from \link{findSwitches}. Related to \link{testSP}
#' and \link{testSC}.
#' @examples
#' \dontrun{
#' igphyml <- "~/apps/igphyml/src/igphyml"
#' data(ExampleAirr)
#' ExampleAirr$sample_id <- sample(ExampleAirr$sample_id)
#' clones <- formatClones(ExampleAirr, trait="sample_id")
#' btrees <- findSwitches(clones[1:2], bootstraps=10, nproc=1,
#' igphyml=igphyml, trait="sample_id")
#' testPS(btrees$switches)
#' }
#' @export
testPS <- function(switches, bylineage=FALSE, pseudocount=0,
alternative=c("less","two.sided","greater")){
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N" &
!!rlang::sym("TO") != !!rlang::sym("FROM") &
!!rlang::sym("FROM") != "UCA")
if(nrow(switches) == 0){
stop("No switches left after filtering!")
}
if(!bylineage){
reps <- switches %>%
dplyr::group_by(!!rlang::sym("REP"), !!rlang::sym("TYPE")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::ungroup() %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("SWITCHES")) %>%
dplyr::mutate(DELTA = !!rlang::sym("RECON") - !!rlang::sym("PERMUTE"))
}else{
reps <- switches %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("CLONE"),
!!rlang::sym("REP")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::ungroup() %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("SWITCHES")) %>%
dplyr::mutate(DELTA = !!rlang::sym("RECON") - !!rlang::sym("PERMUTE"))
}
if(!bylineage){
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}else{
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!rlang::sym("PERMUTE")),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}
means$STAT <- "PS"
reps$STAT <- "PS"
means$REPS <- length(unique(reps$REP))
results <- list()
results$reps <- reps
results$means <- means
results
}
#' Performs SP (switch proportion) test on switch data
#'
#' \code{testSP} performs an SP test
#' @param switches Data frame from findSwitches
#' @param permuteAll Permute among trees?
#' @param from Include only switches from this state?
#' @param to Include only switches to this state?
#' @param dropzeroes Drop switches with zero counts?
#' @param bylineage Perform test for each lineage individually?
#' @param pseudocount Pseudocount for P value calculations
#' @param alternative Perform one-sided (\code{greater} or \code{less})
#' or \code{two.sided} test
#' @param tip_switch maximum tip/switch ratio
#' @param exclude exclude clones with tip/switch ratio > \code{tip_switch}?
#' @return A list containing a \code{tibble} with mean SP statistics, and another
#' with SP statistics per repetition.
#'
#' @details
#' Output data table columns:
#' RECON = SP for observed data
#' PERMUTE = SP for permuted data
#' DELTA = RECON - PERMUTE
#' PLT = p value for DELTA < 0
#' PGT = p value for DELTA < 0
#' \itemize{
#' \item \code{FROM}: State going from.
#' \item \code{TO}: State going to.
#' \item \code{RECON}: SP for observed data.
#' \item \code{PERMUTE}: SP for permuted data.
#' \item \code{DELTA}: RECON - PERMUTE.
#' \item \code{PLT}: p value that DELTA < 0
#' \item \code{PGT}: p value that DELTA > 0
#' \item \code{STAT}: Statistic used (SP).
#' \item \code{REP}: Bootstrap repetition.
#' \item \code{REPS}: Total number of ootstrap repetition.
#'}
#'
#' @seealso Uses output from \link{findSwitches}. Related to \link{testPS}
#' and \link{testSC}.
#' @examples
#' \dontrun{
#' igphyml <- "~/apps/igphyml/src/igphyml"
#' data(ExampleAirr)
#' ExampleAirr$sample_id = sample(ExampleAirr$sample_id)
#' clones = formatClones(ExampleAirr, trait="sample_id")
#' btrees = findSwitches(clones[1:2], bootstraps=10, nproc=1,
#' igphyml=igphyml, trait="sample_id")
#' testSP(btrees$switches)
#' }
#' @export
testSP <- function(switches, permuteAll=FALSE,
from=NULL, to=NULL, dropzeroes=TRUE,
bylineage=FALSE, pseudocount=0, alternative=c("greater","two.sided","less"),
tip_switch=20, exclude=FALSE){
permute <- dplyr::quo(!!rlang::sym("PERMUTE"))
if(permuteAll){
permute <- dplyr::quo(!!rlang::sym("PERMUTEALL"))
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != !!rlang::sym("FROM") &
!!rlang::sym("FROM") != "UCA")
if(!is.null(from)){
from <- dplyr::enquo(from)
switches <- dplyr::filter(switches, !!rlang::sym("FROM") == !!from)
}
if(!is.null(to)){
to <- dplyr::enquo(to)
switches <- dplyr::filter(switches, !!rlang::sym("TO") == !!to)
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N")
if(nrow(switches) == 0){
stop("No switches left after filtering!")
}
if(exclude){
# separate tips
tips <- dplyr::filter(switches, !!rlang::sym("TO")=="NTIP" &
!!rlang::sym("TYPE")=="RECON")
# germline doesn't count in downsampling algorithm
tips$SWITCHES <- tips$SWITCHES - 1
counts <- testPS(switches, bylineage=TRUE)$means
m <- match(counts$CLONE, tips$CLONE)
counts$tips <- tips[m,]$SWITCHES
counts$ratio <- counts$tips/counts$RECON
excluded <- dplyr::filter(counts, !!rlang::sym("ratio") > tip_switch)$CLONE
if(length(excluded) > 0){
warning(paste("Excluding clone(s)",paste(excluded,collapse=",")
,"due to high tip/switch ratio"))
switches <- dplyr::filter(switches,!(!!rlang::sym("CLONE") %in% excluded))
}
}
if(!bylineage){
reps <- switches %>%
dplyr::group_by(!!rlang::sym("REP"), !!rlang::sym("TYPE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES")))
reps <- reps %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("REP")) %>%
dplyr::mutate(PROP = !!rlang::sym("SWITCHES")/
sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("PROP"))
}else{
reps <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("CLONE"),
!!rlang::sym("REP")) %>%
dplyr::mutate(PROP = !!rlang::sym("SWITCHES")/
sum(!!rlang::sym("SWITCHES"))) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("PROP"))
}
reps <- reps %>% dplyr::mutate(DELTA = !!rlang::sym("RECON")-!!permute)
if(dropzeroes){
from_type <-
reps %>% dplyr::group_by(!!rlang::sym("FROM")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("FROM"))
to_type <-
reps %>% dplyr::group_by(!!rlang::sym("TO")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("TO"))
remove <- to_type[to_type %in% from_type]
reps <- reps[!reps$FROM %in% remove &
!reps$TO %in% remove, ]
}
if(sum(is.na(reps$DELTA)) > 0){
na_reps <- unique(reps[is.na(reps$DELTA),]$REP)
warning(paste0("NA delta values in ",
length(na_reps)," replicates, discarding"))
reps <- dplyr::filter(reps, !(!!rlang::sym("REP") %in% na_reps))
}
if(!bylineage){
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}else{
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}
means$STAT <- "SP"
reps$STAT <- "SP"
means$REPS <- length(unique(reps$REP))
results <- list()
results$reps <- reps
results$means <- means
results
}
#' Performs SC (switch count) test on switch data
#'
#' \code{testSC} performs an SC test
#' @param switches Data frame from findSwitches
#' @param permuteAll Permute among trees?
#' @param from Include only switches from this state?
#' @param to Include only switches to this state?
#' @param dropzeroes Drop switches with zero counts?
#' @param bylineage Perform test for each lineage individually?
#' @param pseudocount Pseudocount for P value calculations
#' @param alternative Perform one-sided (\code{greater} or \code{less})
#' or \code{two.sided} test
#' @return A list containing a \code{tibble} with mean SC statistics, and another
#' with SC statistics per repetition.
#'
#' @details
#' Output data table columns:
#' RECON = SC for observed data
#' PERMUTE = SC for permuted data
#' DELTA = RECON - PERMUTE
#' PLT = p value for DELTA < 0
#' PGT = p value for DELTA < 0
#' \itemize{
#' \item \code{FROM}: State going from.
#' \item \code{TO}: State going to.
#' \item \code{RECON}: SC for observed data.
#' \item \code{PERMUTE}: SC for permuted data.
#' \item \code{DELTA}: RECON - PERMUTE.
#' \item \code{PLT}: p value that DELTA < 0
#' \item \code{PGT}: p value that DELTA > 0
#' \item \code{STAT}: Statistic used (SC).
#' \item \code{REP}: Bootstrap repetition.
#' \item \code{REPS}: Total number of ootstrap repetition.
#'}
#'
#' @seealso Uses output from \link{findSwitches}. Related to \link{testPS}
#' and \link{testSP}.
#' @examples
#' \dontrun{
#' igphyml <- "~/apps/igphyml/src/igphyml"
#' data(ExampleAirr)
#' ExampleAirr$sample_id = sample(ExampleAirr$sample_id)
#' clones = formatClones(ExampleAirr, trait="sample_id")
#' btrees = findSwitches(clones[1:2], bootstraps=100, nproc=1,
#' igphyml=igphyml, trait="sample_id", id="temp", dir="temp")
#' testSC(btrees$switches)
#' }
#' @export
testSC <- function(switches,dropzeroes=TRUE,
bylineage=FALSE, pseudocount=0, from=NULL, to=NULL,
permuteAll=FALSE, alternative=c("two.sided","greater","less")){
permute <-dplyr::quo(!!rlang::sym("PERMUTE"))
if(permuteAll){
permute <-dplyr::quo(!!rlang::sym("PERMUTEALL"))
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != !!rlang::sym("FROM") &
!!rlang::sym("FROM") != "UCA")
if(!is.null(from)){
from <-dplyr::enquo(from)
switches <- dplyr::filter(switches, !!rlang::sym("FROM") == !!from)
}
if(!is.null(to)){
to <-dplyr::enquo(to)
switches <- dplyr::filter(switches, !!rlang::sym("TO") == !!to)
}
switches <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N")
if(nrow(switches) == 0){
stop("No switches left after filtering!")
}
if(!bylineage){
reps <- switches %>%
dplyr::group_by(!!rlang::sym("REP"), !!rlang::sym("TYPE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(SWITCHES = sum(!!rlang::sym("SWITCHES")))
reps <- reps %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("REP")) %>%
dplyr::mutate(COUNT = !!rlang::sym("SWITCHES")) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("COUNT"))
}else{
reps <- switches %>%
dplyr::filter(!!rlang::sym("TO") != "N") %>%
dplyr::group_by(!!rlang::sym("TYPE"), !!rlang::sym("CLONE"),
!!rlang::sym("REP")) %>%
dplyr::mutate(COUNT = !!rlang::sym("SWITCHES")) %>%
dplyr::select(-!!rlang::sym("SWITCHES")) %>%
tidyr::spread(!!rlang::sym("TYPE"), !!rlang::sym("COUNT"))
}
reps <- reps %>% dplyr::mutate(DELTA = !!rlang::sym("RECON")-!!permute)
if(dropzeroes){
from_type <-
reps %>% dplyr::group_by(!!rlang::sym("FROM")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("FROM"))
to_type <-
reps %>% dplyr::group_by(!!rlang::sym("TO")) %>%
dplyr::summarize(d = sum(!!rlang::sym("RECON"))) %>%
dplyr::filter(!!rlang::sym("d") == 0) %>%
dplyr::pull(rlang::sym("TO"))
remove <- to_type[to_type %in% from_type]
reps <- reps[!reps$FROM %in% remove &
!reps$TO %in% remove, ]
}
if(!bylineage){
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}else{
if(alternative[1] == "two.sided"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0)*0.5 + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "greater"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PGT = (sum(!!rlang::sym("DELTA") < 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}else if(alternative[1] == "less"){
means <- reps %>%
dplyr::group_by(!!rlang::sym("CLONE"),
!!rlang::sym("FROM"), !!rlang::sym("TO")) %>%
dplyr::summarize(
RECON = mean(!!rlang::sym("RECON")),
PERMUTE = mean(!!permute),
PLT = (sum(!!rlang::sym("DELTA") > 0) +
sum(!!rlang::sym("DELTA") == 0) + pseudocount)/
(dplyr::n() + pseudocount),
DELTA = mean(!!rlang::sym("DELTA")))
}
}
means$STAT <- "SC"
reps$STAT <- "SC"
means$REPS <- length(unique(reps$REP))
results <- list()
results$reps <- reps
results$means <- means
results
}
#' Get divergence from root of tree for each tip
#'
#' \code{getDivergence} get sum of branch lengths leading from the
#' root of the tree. If the germline sequence is included in the tree,
#' this will equal the germline divergence. If germline removed,
#' this will equal the MRCA divergence
#' @param phy Tree object
#' @param minlength Branch lengths to collapse in trees
#' @return A named vector of each tip's divergence from the tree's root.
#'
#' @export
getDivergence = function(phy, minlength=0.001){
tips <- phy$tip.label
phy$edge.length[phy$edge.length < minlength] <- 0
phy <- ape::di2multi(phy,tol=minlength)
uca <- ape::getMRCA(phy,tip=tips)
co <- ape::dist.nodes(phy)
dist <- co[1:length(tips),uca]
names(dist) <- tips
dist
}
#' Resolve polytomies to have the minimum number of single timepoint clades
#'
#' @param phy Tree object
#' @param clone airrClone data object corresponding to \code{phy}
#' @param time Column name holding numeric time information
#' @param sequence Column name holding sequence ID
#' @param germline Germline sequence name
#' @param minlength Branch lengths to collapse in trees
#' @param verbose Print lots of rubbish while running?
#' @return A \code{phylo} tree object in which polytomies are resolved to
#' have the minimum number of single timepoint clades.
#'
#' @details
#' Iteratively identifies polytomies (clusters of < minlength branches),
#' prunes each descendant branch, combines clades with the same timepoint
#' before grouping them back together. Checks to make sure that the divergence
#' of each tip is the same after resolution.
#'
#' @seealso Uses output from \link{getTrees} during \link{correlationTest}.
#' @export
resolvePolytomies = function(phy, clone, minlength=0.001,
time="time", sequence="sequence_id", germline = "Germline",
verbose=FALSE){
data <- clone@data
phy <- rerootTree(ape::di2multi(phy,tol=minlength),
germline=germline,verbose=0)
tips <- phy$tip.label
if(sum(!data[[sequence]] %in% phy$tip.label) > 0){
stop("Tree and data sequence ids don't match")
}
ittybitty_branch = 1e-7
# ape does weird things with zero branches
phy$edge.length[phy$edge.length == 0] = ittybitty_branch
odivergence <- getDivergence(phy,minlength)
polytomies <- table(phy$edge[,1])
polytomies <- names(polytomies[polytomies > 2])
while(length(polytomies) > 0){
if(verbose){
print(polytomies)
}
# for first polytomy in list, remove each subclade
# and record the time of its sequences
target_node <- as.numeric(polytomies[1])
node_index <- which(phy$edge[,1] == target_node)
node_ns <- phy$edge[node_index,2]
edge_ls <- phy$edge.length[node_index]
clades <- list()
times <- list()
for(i in 1:length(node_ns)){
nc <- node_ns[i]
# unfortuantely processing is different if clade is > 1 tip.
if(nc <= length(phy$tip.label)){ #clade is a tip
clades[[i]] <- ape::keep.tip(phy, tip = nc)
times[[i]] <- unique(data[data[[sequence]] %in%
clades[[i]]$tip.label,][[time]])
clades[[i]]$root.edge <- edge_ls[[i]] #store root edge
clades[[i]]$edge.length <- 0
}else{
tips <- ape::extract.clade(phy, node = nc,
collapse.singles=TRUE)$tip.label
clades[[i]] <- ape::keep.tip(phy, tip = tips)
clades[[i]]$root.edge <- edge_ls[[i]] #store root edge
times[[i]] <- unique(data[data[[sequence]] %in%
clades[[i]]$tip.label,][[time]])
if(length(times[[i]]) > 1){
times[[i]] <- "multi"
}
}
}
# combine all clades of each time type in a balanced
# manner, so they remain clades
topclades <- list()
distinct_times <- unique(unlist(times))
for(ctime in distinct_times){
tclades <- clades[unlist(lapply(times,
function(x)x==ctime))]
clade <- tclades[[1]]
if(length(tclades) > 1){
for(i in 2:length(tclades)){
clade <- ape::bind.tree(clade,tclades[[i]],
position=clade$root.edge)
clade <- ape::collapse.singles(clade)
clade$root.edge <- ittybitty_branch
}
}
topclades[[as.character(ctime)]] <- clade
}
# combine top clades into a polytomy
polytomy <- topclades[[1]]
if(length(topclades) > 1){
for(i in 2:length(topclades)){
polytomy <- ape::bind.tree(polytomy,topclades[[i]],
position=polytomy$root.edge)
polytomy <- ape::collapse.singles(polytomy)
polytomy$root.edge <- ittybitty_branch
}
}
tphy <- ape::drop.tip(phy, collapse.singles=TRUE, trim.internal=FALSE, tip =
ape::extract.clade(phy, node = target_node)$tip.label, subtree=TRUE)
pruned_tip <- paste0("[",length(polytomy$tip.label),"_tips]")
ntree <- ape::bind.tree(tphy, polytomy, where=which(tphy$tip.label==pruned_tip),
position=polytomy$root.edge)
ntree <- ape::drop.tip(ntree, tip=pruned_tip)
phy <- ntree
polytomies <- table(phy$edge[,1])
polytomies <- names(polytomies[polytomies > 2])
}
ndivergence <- getDivergence(phy,minlength)
if(length(ndivergence) != length(odivergence)){
stop("Number of tips corrupted during polytomy resolution")
}
if(sum(odivergence[names(ndivergence)] != ndivergence)){
stop("Divergence corrupted during polytomy resolution")
}
phy
}
#' Run correlationTest, based on https://doi.org/10.1111/2041-210X.12466
#'
#' \code{runCorrelationTest} performs root-to-tip regression permutation test
#' @param phy Tree object
#' @param clone airrClone data object corresponding to \code{phy}
#' @param permutations Number of permutations to run
#' @param polyresolve Resolve polytomies to have a minimum number of
#' single timepoint clades
#' @param permutation Permute among single timepoint clades or uniformly
#' among tips
#' @param time Column name holding numeric time information
#' @param sequence Column name holding sequence ID
#' @param germline Germline sequence name
#' @param minlength Branch lengths to collapse in trees
#' @param verbose Print lots of rubbish while running?
#' @param alternative Is alternative that the randomized correlation are greater than
#' or equal to observed, or greater/less than?
#' @return A list of statistics from running the permutation test.
#'
#' @details
#' See \link{correlationTest} for details
#' @seealso \link{correlationTest}.
runCorrelationTest = function(phy, clone, permutations, minlength=0.001,
polyresolve = TRUE, permutation = c("clustered", "uniform"),
time="time", sequence="sequence_id", germline = "Germline",
verbose=TRUE, alternative = c("greater","two.sided")){
data <- clone@data
if(verbose){
print(paste("Analyzing clone: ",clone@clone))
}
if(polyresolve){
if(verbose){
print("resolving polytomies to single timepoint clades")
}
phy <- resolvePolytomies(phy, clone, minlength,
time=time, sequence=sequence, germline = germline,
verbose=verbose)
}
phy <- ape::drop.tip(phy,germline)
tips <- phy$tip.label
if(sum(!data[[sequence]] %in% phy$tip.label) > 0){
stop("Tree and data sequence ids don't match")
}
# dates is named and ordered by tree tips
dates <- data[[time]]
names(dates) <- data[[sequence]]
dates <- dates[tips]
if(dplyr::n_distinct(dates)==1){
warning(paste("Only one timepoint cluster in clone",clone@clone))
results <- list(correlation=NA)
results[["clone"]] <- clone
results[["tree"]] <- phy
results[["random"]] <- NA
results[["random_correlation"]] <- NA
results[["p_gt"]] <- NA
results[["p_lt"]] <- NA
results[["nposs"]] <- NA
results[["nclust"]] <- NA
results[["p"]] <- NA
results[["min_p"]] <- NA
results[["slope"]] <- NA
return(results)
}
divergence <- getDivergence(phy, minlength)
data$divergence <- divergence[data[[sequence]]]
cl <- 1:length(dates) # each tip is its own cluster
if(permutation[1] == "clustered"){
# define all monophyletic clades with > 1 tips
gc <- lapply(ape::subtrees(phy),function(x)x$tip.label)
# identify which clades are single timepoint
single.date.clades <- which(unlist(lapply(gc,function(x)
dplyr::n_distinct(dates[x])))==1)
if(length(single.date.clades)>0){
for(i in 1:length(single.date.clades)){
m <- match(gc[[single.date.clades[i]]],phy$tip.label)
cl[m] <- cl[m[1]] # assign clusters to be the same across single timepoint clades
}
}
cl <- match(cl,unique(cl)) #make clusters 1:(number of clusters)
}
# assign clusters to each sequence
names(cl) <- tips
m <- match(data[[sequence]],names(cl))
data$cluster <- cl[data$sequence_id]
counts <- table(data$cluster,data[[time]])
if(sum(rowSums(counts > 0) > 1) > 0){
stop("Clusters are not single timepoint!")
}
# get ordered list of times for each cluster
ctimes <- unlist(lapply(sort(unique(data$cluster)), function(x)
unique(data[data$cluster == x,][[time]])))
true <- stats::cor(data[[time]],data$divergence)
random <- rep(0, length=permutations)
for(i in 1:permutations){
times <- ctimes[sample(1:length(ctimes))]
data$random_time <- times[data$cluster]
random[i] <- stats::cor(data$divergence, data$random_time)
}
nclust <- dplyr::n_distinct(cl)
nposs <-exp(lfactorial(nclust)-sum(lfactorial(table(ctimes))))
minp <- 1
if(alternative[1] == "two.sided"){
minp <- 0.5
}
clone@data <- data
slope <- summary(stats::lm(data$divergence ~ data[[time]]))$coefficients[2,1]
results <- list(correlation=true)
results[["clone"]] <- clone
results[["tree"]] <- phy
results[["random"]] <- random
results[["random_correlation"]] <- mean(random)
random <- c(random,true)
results[["p_gt"]] <- (sum(true < random) +
sum(true == random)*0.5)/length(random)
results[["p_lt"]] <- (sum(true > random) +
sum(true == random)*0.5)/length(random)
results[["nposs"]] <- nposs
results[["nclust"]] <- nclust
results[["p"]] <- mean(true <= random)
results[["min_p"]] <- max(minp/nposs, minp/(permutations+1))
results[["slope"]] <- slope
results
}
#' Run date randomization test for temporal signal on a set of trees.
#'
#' \code{correlationTest} performs root-to-tip regression date randomization test
#' @param clones A \code{tibble} object containing airrClone and \code{phylo} objects
#' @param permutations Number of permutations to run
#' @param polyresolve Resolve polytomies to have a minimum number of
#' single timepoint clades
#' @param perm_type Permute among single timepoint clades or uniformly
#' among tips
#' @param time Column name holding numeric time information
#' @param sequence Column name holding sequence ID
#' @param germline Germline sequence name
#' @param minlength Branch lengths to collapse in trees
#' @param verbose Print lots of rubbish while running?
#' @param storeTree Store the tree used?
#' @param alternative Is alternative that the randomized correlation are greater than
#' or equal to observed, or greater/less than?
#' @param nproc Number of cores to use for calculations. Parallelizes by tree.
#' @return A \code{tibble} with the same columns as clones, but additional
#' columns corresponding to test statistics for each clone.
#'
#' @details
#' Object returned contains these columns which are added or modified from input:
#' \itemize{
#' \item \code{data}: airrClone object, same as input but with additional columns
#' "cluster" which correspond to permutation cluster, and "divergence."
#' \item \code{slope}: Slope of linear regression between divergence and time.
#' \item \code{correlation}: Correlation between divergence and time.
#' \item \code{p}: p value of correlation compared to permuted correlations.
#' \item \code{random_correlation}: Mean correlation of permutation replicates.
#' \item \code{min_p}: Minimum p value of data, determined by either the number of
#' distinct clade/timepoint combinations or number of permutations.
#' \item \code{nposs}: Number of possible distinct timepoint/clade combinations.
#' \item \code{nclust}: Number of clusters used in permutation. If perm_type="uniform"
#' this is the number of tips.
#' \item \code{p_gt/p_lt}: P value that permuted correlations are greater or less
#' than observed correlation. Only returned if alternative = "two.sided"
#' \item \code{test_trees}: The \link{phylo} tree objects used, possibly with
#' resolved polytomies.
#' }
#' @seealso Uses output from \code{getTrees}.
#' @export
correlationTest = function(clones, permutations=1000, minlength=0.001,
perm_type = c("clustered", "uniform"), time="time",
sequence="sequence_id", germline = "Germline",
verbose=FALSE, polyresolve = TRUE,
alternative = c("greater","two.sided"),
storeTree = FALSE, nproc=1){
if(!"tbl" %in% class(clones)){
print(paste("clones is of class",class(clones)))
stop("clones must be a tibble of airrClone objects!")
}else{
if(!inherits(clones$data[[1]], "airrClone")){
print(paste("clones is list of class",class(clones$data[[1]])))
stop("clones must be a list of airrClone objects!")
}
}
if(!"trees" %in% names(clones)){
stop("clones must have trees column!")
}
time_check <- unlist(lapply(clones$data, function(x)!time %in% names(x@data)))
if(sum(time_check) > 0){
stop(paste("Time column",time,"not found in clone object (must be trait value in formatClones)"))
}
time_check <- unlist(lapply(clones$data, function(x)!is.numeric(x@data[[time]])))
if(sum(time_check) > 0){
stop(paste("Time column",time,"contains non-numeric values, impossible to continue"))
}
results <- parallel::mclapply(1:nrow(clones),function(x)
runCorrelationTest(clones$trees[[x]], clones$data[[x]],
permutations, minlength=minlength, polyresolve=polyresolve,
permutation=perm_type, time= time,
sequence=sequence, germline=germline,
verbose=verbose, alternative=alternative),
mc.cores=nproc)
clones$data <- lapply(results,function(x)x$clone)
if(storeTree){
clones$test_trees <- lapply(results,function(x)x$tree)
}
cols <- c("slope", "p", "correlation", "random_correlation",
"min_p", "nposs", "nclust")
if(alternative[1] == "two.sided"){
cols <- c(cols, c("p_gt", "p_lt"))
}
for(col in cols){
clones[[col]] <- unlist(lapply(results,function(x)x[[col]]))
}
clones
}
#' Finds the Robinson-Fould's cluster distance between phylogenies.
#'
#' \code{calcRF} Calculates the RF distance between two phylogenetic trees with
#' the same tips and tip labels.
#' @param tree_1 A \code{phylo} object
#' @param tree_2 A \code{phylo} object
#' @param nproc Number of cores to use for calculations.
#'
#' @return The RF cluster value for the two input trees.
#'
#' @export
calcRF <- function(tree_1, tree_2, nproc = 1){
tip_amount_check <- length(tree_1$tip.label) == length(tree_2$tip.label)
if(!tip_amount_check){
stop("trees do not have the same number of tips")
}
tip_check <- dplyr::setdiff(tree_1$tip.label, tree_2$tip.label)
# change this
if(!identical(tip_check, character(0))){
stop("tree tip labels are not identical")
}
tree_1_df <- splits_func(list(tree_1),1)
tree_2_df <- splits_func(list(tree_2), 1)
total_mismatches <- unlist(parallel::mclapply(1:nrow(tree_1_df), function(x){
tree_1_sub <- tree_1_df$found[[x]]
mismatch_vector <- unlist(lapply(1:nrow(tree_2_df), function(y){
mismatches_1 <- dplyr::setdiff(tree_2_df$found[[y]], tree_1_sub)
mismatches_2 <- dplyr::setdiff(tree_1_sub, tree_2_df$found[[y]])
if(identical(mismatches_1, character(0)) & identical(mismatches_2, character(0))){
value <- "match"
} else{
value <- "mismatch"
}
return(value)
}))
if("match" %in% mismatch_vector){
tobind <- 0
} else{
tobind <- 1
}
return(tobind)
}, mc.cores = nproc))
total_mismatches <- sum(total_mismatches)
clone_mismatches <- unlist(parallel::mclapply(1:nrow(tree_2_df), function(x){
tree_2_sub <- tree_2_df$found[[x]]
mismatch_vector <- unlist(lapply(1:nrow(tree_1_df), function(y){
mismatches_1 <- setdiff(tree_1_df$found[[y]], tree_2_sub)
mismatches_2 <- setdiff(tree_2_sub, tree_1_df$found[[y]])
if(identical(mismatches_1, character(0)) & identical(mismatches_2, character(0))){
value <- "match"
} else{
value <- "mismatch"
}
return(value)
}))
if("match" %in% mismatch_vector){
tobind <- 0
} else{
tobind <- 1
}
return(tobind)
}, mc.cores = nproc))
clone_mismatches <- sum(clone_mismatches)
all_mismatches <- total_mismatches + clone_mismatches
return(all_mismatches)
}
Try the dowser package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.