R/internal-funcs.R

Defines functions graph.posterior.marginal attractor.posterior.2 attractor.2.detail attractor.2 attractor.distance.summary apply.op score.attractor summarize.attractor.bool summarize.attractor.2 summarize.attractor.1 summarize.attractor.calc condense.attractor f4p ArrayToHash read.cc.app.posterior.bool read.cc.app.posterior read.cc.app.bool read.cc.app.table.version read.cc.app

########## for R library

####################################################################

read.cc.app <- function(fname) {
  junk <- scan(fname, nlines = 5, what = "character", sep = "\n")

  StartDate <- junk[1]
  EndDate <- junk[2]
  InputFile <- junk[3]
  OutputFile <- junk[4]
  Memo <- junk[5]

  junk <- scan(fname, skip = 5)

  PerturbType <- junk[1]
  ScoreType <- junk[2]
  nBackupStage <- junk[3]
  MaxStage <- junk[4]
  MaxTransition <- junk[5]
  epsilon <- junk[6]
  beta <- junk[7]
  chi0 <- junk[8]
  delta <- junk[9]
  ne <- junk[10]
  m0 <- junk[11]
  MaxDegree <- junk[12]
  PAddParent <- junk[13]
  PExchangeParent <- junk[14]
  NeighborDegree <- junk[15]
  icount <- 15
  pNeighborhood <- NULL
  if (NeighborDegree > 1) {
    for (i in seq_len((NeighborDegree - 1))) {
      icount <- icount + 1
      pNeighborhood <- c(pNeighborhood, junk[icount])
    }
  }
  icount <- icount + 1
  rho <- junk[icount]
  icount <- icount + 1
  xSeed <- junk[icount]
  icount <- icount + 1
  nGene <- junk[icount]
  icount <- icount + 1
  nExp <- junk[icount]
  icount <- icount + 1
  edgePenalty <- junk[icount]

  DataResponse <- matrix(junk[c(1:(nGene * nExp)) + icount], nGene, nExp, byrow = TRUE)
  icount <- icount + (nGene * nExp)
  DataPerturbation <- matrix(junk[c(1:(nGene * nExp)) + icount], nGene, nExp, byrow = TRUE)
  icount <- icount + (nGene * nExp)

  icount <- icount + 1
  NewScore <- junk[icount]
  icount <- icount + 1
  MinScore <- junk[icount]

  Fit.InDegree <- junk[(1:nGene) + icount]
  icount <- icount + nGene
  net.graph <- NULL
  for (j in seq_len(nGene)) {
    if (Fit.InDegree[j] > 0) {
      incr <- Fit.InDegree[j]
      net.graph <- append(net.graph, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    if (Fit.InDegree[j] == 0) {
      net.graph <- append(net.graph, list(NULL))
    }
  }
  net.op <- NULL
  for (j in seq_len(nGene)) {
    incr <- 3^Fit.InDegree[j]
    net.op <- append(net.op, list(junk[c(1:incr) + icount]))
    icount <- icount + incr
  }
  net.table <- NULL

  MinFit.InDegree <- junk[(1:nGene) + icount]
  icount <- icount + nGene
  min.net.graph <- NULL
  for (j in seq_len(nGene)) {
    if (MinFit.InDegree[j] > 0) {
      incr <- MinFit.InDegree[j]
      min.net.graph <- append(min.net.graph, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    if (MinFit.InDegree[j] == 0) {
      min.net.graph <- append(min.net.graph, list(NULL))
    }
  }
  min.net.op <- NULL
  for (j in seq_len(nGene)) {
    incr <- 3^MinFit.InDegree[j]
    min.net.op <- append(min.net.op, list(junk[c(1:incr) + icount]))
    icount <- icount + incr
  }


  icount <- icount + 1
  FinalTemperature <- junk[icount]
  icount <- icount + 1
  nStage <- junk[icount]

  MuTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  SigmaTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  SigmaRhoTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  TemperatureTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  AcceptanceTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage

  fit.obj <- list(
    xSeed = xSeed,
    nStage = nStage,
    PerturbType = PerturbType,
    ScoreType = ScoreType,
    nBackupStage = nBackupStage,
    nExp = nExp,
    nGene = nGene,
    edgePenalty = edgePenalty,
    MaxStage = MaxStage,
    MaxTransition = MaxTransition,
    epsilon = epsilon,
    beta = beta,
    chi0 = chi0,
    delta = delta,
    ne = ne,
    m0 = m0,
    m1 = NULL,
    m2 = NULL,
    MaxDegree = MaxDegree,
    PAddParent = PAddParent,
    PExchangeParent = PExchangeParent,
    NeighborDegree = NeighborDegree,
    pNeighborhood = pNeighborhood,
    rho = rho,
    FinalTemperature = FinalTemperature,
    NewScore = NewScore,
    MinScore = MinScore,
    MinStage = NULL,
    MinTransition = NULL,
    MinFit.InDegree = MinFit.InDegree,
    Fit.InDegree = Fit.InDegree,
    min.net.graph = min.net.graph,
    min.net.op = min.net.op,
    net.graph = net.graph,
    net.op = net.op,
    MuTrace = MuTrace,
    SigmaTrace = SigmaTrace,
    SigmaRhoTrace = SigmaRhoTrace,
    TemperatureTrace = TemperatureTrace,
    AcceptanceTrace = AcceptanceTrace,
    DataResponse = DataResponse,
    DataPerturbation = DataPerturbation,
    Memo = Memo,
    StartDate = StartDate,
    EndDate = EndDate,
    InputFile = InputFile,
    OutputFile = OutputFile
  )

  return(fit.obj)
}

###################################################


read.cc.app.table.version <- function(fname) {
  junk <- scan(fname, nlines = 5, what = "character", sep = "\n")

  StartDate <- junk[1]
  EndDate <- junk[2]
  InputFile <- junk[3]
  OutputFile <- junk[4]
  Memo <- junk[5]

  junk <- scan(fname, skip = 5)

  PerturbType <- junk[1]
  ScoreType <- junk[2]
  nBackupStage <- junk[3]
  MaxStage <- junk[4]
  MaxTransition <- junk[5]
  epsilon <- junk[6]
  beta <- junk[7]
  chi0 <- junk[8]
  delta <- junk[9]
  ne <- junk[10]
  m0 <- junk[11]
  MaxDegree <- junk[12]
  PAddParent <- junk[13]
  PRemoveParent <- junk[14]
  PExchangeParent <- junk[15]
  NeighborDegree <- junk[16]
  icount <- 16
  pNeighborhood <- NULL
  if (NeighborDegree > 1) {
    for (i in seq_len((NeighborDegree - 1))) {
      icount <- icount + 1
      pNeighborhood <- c(pNeighborhood, junk[icount])
    }
  }
  icount <- icount + 1
  rho <- junk[icount]
  icount <- icount + 1
  xSeed <- junk[icount]
  icount <- icount + 1
  nGene <- junk[icount]
  icount <- icount + 1
  nExp <- junk[icount]
  icount <- icount + 1
  edgePenalty <- junk[icount]

  DataResponse <- matrix(junk[c(1:(nGene * nExp)) + icount], nGene, nExp, byrow = TRUE)
  icount <- icount + (nGene * nExp)
  DataPerturbation <- matrix(junk[c(1:(nGene * nExp)) + icount], nGene, nExp, byrow = TRUE)
  icount <- icount + (nGene * nExp)

  icount <- icount + 1
  NewScore <- junk[icount]
  icount <- icount + 1
  MinScore <- junk[icount]

  Fit.InDegree <- junk[(1:nGene) + icount]
  icount <- icount + nGene
  net.graph <- NULL
  for (j in seq_len(nGene)) {
    if (Fit.InDegree[j] > 0) {
      incr <- Fit.InDegree[j]
      net.graph <- append(net.graph, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    if (Fit.InDegree[j] == 0) {
      net.graph <- append(net.graph, list(NULL))
    }
  }
  net.op <- NULL
  for (j in seq_len(nGene)) {
    incr <- 3^Fit.InDegree[j]
    net.op <- append(net.op, list(junk[c(1:incr) + icount]))
    icount <- icount + incr
  }
  net.table <- NULL

  MinFit.InDegree <- junk[(1:nGene) + icount]
  icount <- icount + nGene
  min.net.graph <- NULL
  for (j in seq_len(nGene)) {
    if (MinFit.InDegree[j] > 0) {
      incr <- MinFit.InDegree[j]
      min.net.graph <- append(min.net.graph, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    if (MinFit.InDegree[j] == 0) {
      min.net.graph <- append(min.net.graph, list(NULL))
    }
  }
  min.net.op <- NULL
  for (j in seq_len(nGene)) {
    incr <- 3^MinFit.InDegree[j]
    min.net.op <- append(min.net.op, list(junk[c(1:incr) + icount]))
    icount <- icount + incr
  }


  icount <- icount + 1
  FinalTemperature <- junk[icount]
  icount <- icount + 1
  nStage <- junk[icount]

  MuTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  SigmaTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  SigmaRhoTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  TemperatureTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage

  fit.obj <- list(
    xSeed = xSeed,
    nStage = nStage,
    PerturbType = PerturbType,
    ScoreType = ScoreType,
    nBackupStage = nBackupStage,
    nExp = nExp,
    nGene = nGene,
    edgePenalty = edgePenalty,
    MaxStage = MaxStage,
    MaxTransition = MaxTransition,
    epsilon = epsilon,
    beta = beta,
    chi0 = chi0,
    delta = delta,
    ne = ne,
    m0 = m0,
    m1 = NULL,
    m2 = NULL,
    MaxDegree = MaxDegree,
    PAddParent = PAddParent,
    PRemoveParent = PRemoveParent,
    PExchangeParent = PExchangeParent,
    NeighborDegree = NeighborDegree,
    pNeighborhood = pNeighborhood,
    rho = rho,
    FinalTemperature = FinalTemperature,
    NewScore = NewScore,
    MinScore = MinScore,
    MinStage = NULL,
    MinTransition = NULL,
    MinFit.InDegree = MinFit.InDegree,
    Fit.InDegree = Fit.InDegree,
    min.net.graph = min.net.graph,
    min.net.op = min.net.op,
    net.graph = net.graph,
    net.op = net.op,
    MuTrace = MuTrace,
    SigmaTrace = SigmaTrace,
    SigmaRhoTrace = SigmaRhoTrace,
    TemperatureTrace = TemperatureTrace,
    AcceptanceTrace = NULL,
    DataResponse = DataResponse,
    DataPerturbation = DataPerturbation,
    Memo = Memo,
    StartDate = StartDate,
    EndDate = EndDate,
    InputFile = InputFile,
    OutputFile = OutputFile
  )

  return(fit.obj)
}



###################################################

read.cc.app.bool <- function(fname) {
  junk <- scan(fname, nlines = 5, what = "character", sep = "\n")

  StartDate <- junk[1]
  EndDate <- junk[2]
  InputFile <- junk[3]
  OutputFile <- junk[4]
  Memo <- junk[5]

  junk <- scan(fname, skip = 5)

  PerturbType <- junk[1]
  ScoreType <- junk[2]
  nBackupStage <- junk[3]
  MaxStage <- junk[4]
  MaxTransition <- junk[5]
  epsilon <- junk[6]
  beta <- junk[7]
  chi0 <- junk[8]
  delta <- junk[9]
  ne <- junk[10]
  m0 <- junk[11]
  MaxDegree <- junk[12]
  PAddParent <- junk[13]
  PExchangeParent <- junk[14]
  NeighborDegree <- junk[15]
  icount <- 15
  pNeighborhood <- NULL
  if (NeighborDegree > 1) {
    for (i in seq_len((NeighborDegree - 1))) {
      icount <- icount + 1
      pNeighborhood <- c(pNeighborhood, junk[icount])
    }
  }
  icount <- icount + 1
  rho <- junk[icount]
  icount <- icount + 1
  xSeed <- junk[icount]
  icount <- icount + 1
  nGene <- junk[icount]
  icount <- icount + 1
  nExp <- junk[icount]
  icount <- icount + 1
  edgePenalty <- junk[icount]
  icount <- icount + 1
  cyclePenaltyN <- junk[icount]
  cyclePenaltyKnots <- junk[icount + c(1:(2 * cyclePenaltyN + 2))]
  icount <- icount + (2 * cyclePenaltyN + 2)

  DataResponse <- matrix(junk[c(1:(nGene * nExp)) + icount], nGene, nExp, byrow = TRUE)
  icount <- icount + (nGene * nExp)
  DataPerturbation <- matrix(junk[c(1:(nGene * nExp)) + icount], nGene, nExp, byrow = TRUE)
  icount <- icount + (nGene * nExp)

  icount <- icount + 1
  NewScore <- junk[icount]
  icount <- icount + 1
  MinScore <- junk[icount]

  Fit.InDegree <- junk[(1:nGene) + icount]
  icount <- icount + nGene
  net.graph <- NULL
  for (j in seq_len(nGene)) {
    if (Fit.InDegree[j] > 0) {
      incr <- Fit.InDegree[j]
      net.graph <- append(net.graph, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    if (Fit.InDegree[j] == 0) {
      net.graph <- append(net.graph, list(NULL))
    }
  }
  net.op <- NULL
  for (j in seq_len(nGene)) {
    incr <- 2^Fit.InDegree[j]
    net.op <- append(net.op, list(junk[c(1:incr) + icount]))
    icount <- icount + incr
  }
  net.table <- NULL

  MinFit.InDegree <- junk[(1:nGene) + icount]
  icount <- icount + nGene
  min.net.graph <- NULL
  for (j in seq_len(nGene)) {
    if (MinFit.InDegree[j] > 0) {
      incr <- MinFit.InDegree[j]
      min.net.graph <- append(min.net.graph, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    if (MinFit.InDegree[j] == 0) {
      min.net.graph <- append(min.net.graph, list(NULL))
    }
  }
  min.net.op <- NULL
  for (j in seq_len(nGene)) {
    incr <- 2^MinFit.InDegree[j]
    min.net.op <- append(min.net.op, list(junk[c(1:incr) + icount]))
    icount <- icount + incr
  }


  icount <- icount + 1
  FinalTemperature <- junk[icount]
  icount <- icount + 1
  nStage <- junk[icount]

  MuTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  SigmaTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  SigmaRhoTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage
  TemperatureTrace <- junk[c(1:nStage) + icount]
  icount <- icount + nStage

  fit.obj <- list(
    xSeed = xSeed,
    nStage = nStage,
    PerturbType = PerturbType,
    ScoreType = ScoreType,
    nBackupStage = nBackupStage,
    nExp = nExp,
    nGene = nGene,
    edgePenalty = edgePenalty,
    cyclePenaltyN = cyclePenaltyN,
    cyclePenaltyKnots = cyclePenaltyKnots,
    MaxStage = MaxStage,
    MaxTransition = MaxTransition,
    epsilon = epsilon,
    beta = beta,
    chi0 = chi0,
    delta = delta,
    ne = ne,
    m0 = m0,
    m1 = NULL,
    m2 = NULL,
    MaxDegree = MaxDegree,
    PAddParent = PAddParent,
    PExchangeParent = PExchangeParent,
    NeighborDegree = NeighborDegree,
    pNeighborhood = pNeighborhood,
    rho = rho,
    FinalTemperature = FinalTemperature,
    NewScore = NewScore,
    MinScore = MinScore,
    MinStage = NULL,
    MinTransition = NULL,
    MinFit.InDegree = MinFit.InDegree,
    Fit.InDegree = Fit.InDegree,
    min.net.graph = min.net.graph,
    min.net.op = min.net.op,
    net.graph = net.graph,
    net.op = net.op,
    MuTrace = MuTrace,
    SigmaTrace = SigmaTrace,
    SigmaRhoTrace = SigmaRhoTrace,
    TemperatureTrace = TemperatureTrace,
    AcceptanceTrace = NULL,
    DataResponse = DataResponse,
    DataPerturbation = DataPerturbation,
    Memo = Memo,
    StartDate = StartDate,
    EndDate = EndDate,
    InputFile = InputFile,
    OutputFile = OutputFile
  )

  return(fit.obj)
}



########################################################

read.cc.app.posterior <- function(fname) {
  junk <- scan(fname, nlines = 5, what = "character", sep = "\n")

  StartDate <- junk[1]
  EndDate <- junk[2]
  FitFile <- junk[3]
  OutputPosteriorFile <- junk[4]
  OutputPosteriorMemo <- junk[5]

  junk <- scan(fname, skip = 5)

  m0 <- junk[1]
  mdelta <- junk[2]
  parameter <- junk[3]
  xSeed <- junk[4]
  FinalTemperature <- junk[5]
  nGene <- junk[6]
  icount <- 6

  fit.score <- rep(NA, m0)
  fit.indegree.list <- NULL
  for (i in seq_len(m0)) {
    fit.indegree.list <- append(fit.indegree.list, list(NULL))
  }
  fit.graph.list <- fit.indegree.list
  fit.op.list <- fit.indegree.list

  for (iii in seq_len(m0)) {
    icount <- icount + 1
    fit.score[iii] <- junk[icount]

    fit.indegree.list[[iii]] <- junk[(1:nGene) + icount]
    icount <- icount + nGene

    fit.graph <- NULL
    for (j in seq_len(nGene)) {
      if (fit.indegree.list[[iii]][j] > 0) {
        incr <- fit.indegree.list[[iii]][j]
        fit.graph <- append(fit.graph, list(junk[c(1:incr) + icount]))
        icount <- icount + incr
      }
      if (fit.indegree.list[[iii]][j] == 0) {
        fit.graph <- append(fit.graph, list(NULL))
      }
    }
    fit.graph.list[[iii]] <- fit.graph
    fit.op <- NULL
    for (j in seq_len(nGene)) {
      incr <- 3^fit.indegree.list[[iii]][j]
      fit.op <- append(fit.op, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    fit.op.list[[iii]] <- fit.op
  }

  posterior.obj <- list(
    StartDate = StartDate,
    EndDate = EndDate,
    FitFile = FitFile,
    OutputPosteriorFile = OutputPosteriorFile,
    OutputPosteriorMemo = OutputPosteriorMemo,
    m0 = m0,
    mdelta = mdelta,
    parameter = parameter,
    xSeed = xSeed,
    FinalTemperature = FinalTemperature,
    nGene = nGene,
    fit.score = fit.score,
    fit.indegree.list = fit.indegree.list,
    fit.graph.list = fit.graph.list,
    fit.op.list = fit.op.list
  )

  return(posterior.obj)
}

########################################################

read.cc.app.posterior.bool <- function(fname) {
  junk <- scan(fname, nlines = 5, what = "character", sep = "\n")

  StartDate <- junk[1]
  EndDate <- junk[2]
  FitFile <- junk[3]
  OutputPosteriorFile <- junk[4]
  OutputPosteriorMemo <- junk[5]

  junk <- scan(fname, skip = 5)

  m0 <- junk[1]
  mdelta <- junk[2]
  parameter <- junk[3]
  xSeed <- junk[4]
  FinalTemperature <- junk[5]
  nGene <- junk[6]
  icount <- 6

  fit.score <- rep(NA, m0)
  fit.indegree.list <- NULL
  for (i in seq_len(m0)) {
    fit.indegree.list <- append(fit.indegree.list, list(NULL))
  }
  fit.graph.list <- fit.indegree.list
  fit.op.list <- fit.indegree.list

  for (iii in seq_len(m0)) {
    icount <- icount + 1
    fit.score[iii] <- junk[icount]

    fit.indegree.list[[iii]] <- junk[(1:nGene) + icount]
    icount <- icount + nGene

    fit.graph <- NULL
    for (j in seq_len(nGene)) {
      if (fit.indegree.list[[iii]][j] > 0) {
        incr <- fit.indegree.list[[iii]][j]
        fit.graph <- append(fit.graph, list(junk[c(1:incr) + icount]))
        icount <- icount + incr
      }
      if (fit.indegree.list[[iii]][j] == 0) {
        fit.graph <- append(fit.graph, list(NULL))
      }
    }
    fit.graph.list[[iii]] <- fit.graph
    fit.op <- NULL
    for (j in seq_len(nGene)) {
      incr <- 2^fit.indegree.list[[iii]][j]
      fit.op <- append(fit.op, list(junk[c(1:incr) + icount]))
      icount <- icount + incr
    }
    fit.op.list[[iii]] <- fit.op
  }

  posterior.obj <- list(
    StartDate = StartDate,
    EndDate = EndDate,
    FitFile = FitFile,
    OutputPosteriorFile = OutputPosteriorFile,
    OutputPosteriorMemo = OutputPosteriorMemo,
    m0 = m0,
    mdelta = mdelta,
    parameter = parameter,
    xSeed = xSeed,
    FinalTemperature = FinalTemperature,
    nGene = nGene,
    fit.score = fit.score,
    fit.indegree.list = fit.indegree.list,
    fit.graph.list = fit.graph.list,
    fit.op.list = fit.op.list
  )

  return(posterior.obj)
}

####################################################################

ArrayToHash <- function(pIA, nOutcomes = 3) {
  n <- length(pIA)
  tempi <- 0

  for (i in seq_len(n)) {
    tempi <- tempi + (nOutcomes^(i - 1)) * (pIA[i] - 1)
  }

  tempi <- tempi + 1

  return(tempi)
}

############################################################

f4p <- function(x) {
  sum(4^x)
}

############################################################

condense.attractor <- function(m, thr = 0) {
  nr <- dim(m)[1]
  nc <- dim(m)[2]
  mm <- m


  for (i in seq_len(nc)) {
    mm[, i][m[, i] == 9] <- 2
  }
  mm <- mm + 1
  code <- apply(mm, 1, f4p)

  unicode <- unique(code)

  p.attr <- rep(0, length(unicode))
  uni.attr <- matrix(0, length(unicode), nc)

  for (i in seq_len(length(unicode))) {
    p.attr[i] <- mean(code == unicode[i])
    ind <- min(c(1:nr)[code == unicode[i]])
    uni.attr[i, ] <- m[ind, ]
  }

  ind <- rev(sort.list(p.attr))
  uni.attr <- as.matrix(uni.attr[ind, ])
  p.attr <- p.attr[ind]
  uni.attr <- uni.attr[p.attr >= thr, ]
  p.attr <- p.attr[p.attr >= thr]


  return(list(p.attr = p.attr, uni.attr = uni.attr))
}

############################################################

summarize.attractor.calc <- function(m01) {
  m0 <- m01[1]
  m1 <- m01[2]

  ans <- 0
  if ((m0 == -1) & (m1 <= 0)) {
    ans <- -1
  }
  if ((m0 == -1) & (m1 == 1)) {
    ans <- 9
  }
  if ((m0 >= 0) & (m1 == 1)) {
    ans <- 1
  }

  return(ans)
}

############################################################


summarize.attractor.1 <- function(m) {
  m <- as.matrix(m)
  attractor <- apply(m, 1, mean)
  return(attractor)
}

############################################################


summarize.attractor.2 <- function(m) {
  m <- as.matrix(m)
  m0 <- apply(m, 1, min)
  m1 <- apply(m, 1, max)
  attractor <- apply(cbind(m0, m1), 1, summarize.attractor.calc)
  return(attractor)
}


############################################################



summarize.attractor.bool <- function(m) {
  m <- as.matrix(m)
  m0 <- apply(m, 1, min)
  m1 <- apply(m, 1, max)
  attractor <- 1 * (m0 == 1 & m1 == 1) + 9 * (m0 < m1)
  return(attractor)
}


############################################################


score.attractor <- function(x) {
  obs <- x[1]
  mn <- x[2]
  mx <- x[3]

  ans <- 1
  if ((obs == 1) & (mn == 1) & (mx <= 2)) {
    ans <- 0
  }
  if ((obs == 2) & (mn == 2) & (mx == 2)) {
    ans <- 0
  }
  if ((obs == 3) & (mn >= 2) & (mx == 3)) {
    ans <- 0
  }

  return(ans)
}


############################################################

apply.op <- function(vec, tn.model) {
  n <- length(vec)
  new.vec <- vec
  for (i in seq_len(n)) {
    if (tn.model$degree[i] == 0) {
      new.vec[i] <- 2
    }
    if (tn.model$degree[i] > 0) {
      in.state <- vec[tn.model$graph[1:tn.model$degree[i], i]]
      tempi <- ArrayToHash(in.state, 3)
      new.vec[i] <- tn.model$table[tempi, i]
    }
  }
  return(new.vec)
}

############################################################

attractor.distance.summary <- function(vec0, tn.model, wt, forced.genes = NULL, forced.states = NULL, nOutcomes = 3) {
  vec.list <- cbind(NULL, vec0)

  list.pos <- 0
  list.size <- 1
  while (list.pos == 0) {
    new.vec <- apply.op(vec.list[, list.size], tn.model)
    if (!wt) {
      new.vec[forced.genes] <- forced.states
    }
    for (i in seq_len(list.size)) {
      if (sum(new.vec != vec.list[, i]) == 0) {
        list.pos <- i
      }
    }
    vec.list <- cbind(vec.list, new.vec)
    list.size <- list.size + 1
  }


  attractor <- cbind(vec0, vec.list[, (list.pos + 1):list.size]) - nOutcomes + 1
  return(list(attractor = attractor, path = vec.list - nOutcomes + 1))
}


#############################################################################################
##
## visible functions
##
#############################################################################################


attractor.2 <- function(tn.model, model.obj, wt, nOutcomes = 3) {
  n.gene <- dim(model.obj$initial.states)[1]
  n.sample <- dim(model.obj$initial.states)[2]

  attractor.matrix <- matrix(NA, n.gene, n.sample)

  for (i in seq_len(n.sample)) {
    fg <- model.obj$forced.gene.list[[i]]$forced.genes
    fs <- model.obj$forced.gene.list[[i]]$forced.states
    # vec0<-model.obj$initial.states[,i]
    vec0 <- rep(2, n.gene)
    vec0[fg] <- fs
    junk <- attractor.distance.summary(vec0, tn.model, wt, fg, fs)
    nc <- dim(junk$attractor)[2]
    if (nOutcomes == 3) {
      attractor.matrix[, i] <- summarize.attractor.2(junk$attractor[, 2:nc])
    }
    if (nOutcomes == 2) {
      attractor.matrix[, i] <- summarize.attractor.bool(junk$attractor[, 2:nc])
    }
  }

  return(attractor.matrix)
}

#######################################

attractor.2.detail <- function(net.op, net.graph, model.obj, wt) {
  n.gene <- dim(model.obj$initial.states)[1]
  n.sample <- dim(model.obj$initial.states)[2]

  junk.list <- NULL

  for (i in seq_len(n.sample)) {
    fg <- model.obj$forced.gene.list[[i]]$forced.genes
    fs <- model.obj$forced.gene.list[[i]]$forced.states
    vec0 <- model.obj$initial.states[, i]
    junk <- attractor.distance.summary(vec0, model.obj, wt, fg, fs, model.obj$nOutcomes)
    junk.list <- append(junk.list, list(junk))
  }
  return(junk.list)
}

#####################################

attractor.posterior.2 <- function(posteriorObj, model.obj, wt) {
  n.post <- dim(tableObjs(posteriorObj))[3]
  n.gene <- dim(perturbationObj(posteriorObj))[1]
  n.sample <- dim(perturbationObj(posteriorObj))[2]

  attractor.list <- NULL
  for (i in seq_len(n.sample)) {
    attractor.list <- append(attractor.list, list(NULL))
  }

  for (i in seq_len(n.sample)) {
    attractor.list[[i]] <- matrix(0, n.post, n.gene)
  }

  for (i in seq_len(n.post)) {
    tn.model <- list(
      table = tableObjs(posteriorObj)[, , i],
      graph = graphObjs(posteriorObj)[, , i],
      degree = degreeObjs(posteriorObj)[i, ]
    )

    junk <- attractor.2(tn.model, model.obj, wt)
    for (j in seq_len(n.sample)) {
      attractor.list[[j]][i, ] <- junk[, j]
    }
  }

  return(attractor.list)
}

####################################################################


graph.posterior.marginal <- function(posteriorObj) {
  n.post <- dim(tableObjs(posteriorObj))[3]
  n.gene <- dim(perturbationObj(posteriorObj))[1]
  n.sample <- dim(perturbationObj(posteriorObj))[2]


  parent.post.table <- matrix(0, n.gene, n.gene + 1)

  for (i in seq_len(n.gene)) {
    for (j in seq_len(n.post)) {
      gr <- graphObjs(posteriorObj)[, , j]

      pp <- c(gr[, i], 0, 0)[1:2]

      if (pp[1] == 0 & pp[2] == 0) {
        parent.post.table[i, pp[1] + 1] <- parent.post.table[i, pp[1] + 1] + 1
      }
      if (pp[1] + pp[2] > 0) {
        parent.post.table[i, pp[1] + 1] <- parent.post.table[i, pp[1] + 1] + 1
        parent.post.table[i, pp[2] + 1] <- parent.post.table[i, pp[2] + 1] + 1
      }
    }
  }

  parent.post.table <- parent.post.table / n.post

  return(parent.post.table)
}

####################################################################
mccallm/ternarynet documentation built on Feb. 26, 2024, 3:51 a.m.