Functional Labeled Optimal Partitioning

Description

Main function for computing optimal segmentation model with Poisson loss, up-down constraints, and label constraints.

Usage

FLOPART(coverage, label, penalty)

Arguments

Value

list with named elements: coverage_dt is a data table with columns chromStart, chromEnd, count, weight; label_dt is a data table with columns chromStart, chromEnd, annotation, type, firstRow, lastRow; cost_mat is a Nx2 numeric matrix of optimal penalized Poisson loss values up to each data point and in each state; intervals_mat is a Nx2 integer matrix of counts of intervals used to store the optimal cost function, useful for analyzing time/space complexity; segments_dt is a data table with columns chromStart, chromEnd, status, mean.

Author(s)

Toby Dylan Hocking

Examples

library(data.table)
data("Mono27ac.simple", package="FLOPART")
Mono27ac.simple
label.pen <- 1400
fit <- with(Mono27ac.simple, FLOPART::FLOPART(coverage, label, label.pen))
lapply(fit, head)

## Plot data and model.
ann.colors <- c(
  noPeaks="orange",
  peakStart="#efafaf",
  peakEnd="#ff4c4c")
model.color <- "blue"
(peaks.dt <- fit[["segments_dt"]][status=="peak"][, peak.y := -2][])
if(require("ggplot2")){
  ggplot()+
    ggtitle("Model with label constraints (FLOPART)")+
    scale_fill_manual("label", values=ann.colors)+
    geom_rect(aes(
      xmin=chromStart, xmax=chromEnd,
      ymin=-Inf, ymax=Inf,
      fill=annotation),
      alpha=0.5,
      color="grey",
      data=Mono27ac.simple[["label"]])+
    geom_step(aes(
      chromStart, count),
      data=Mono27ac.simple[["coverage"]],
      color="grey50")+
    geom_step(aes(
      chromStart, mean),
      data=fit[["segments_dt"]],
      color=model.color)+
    geom_segment(aes(
      chromStart, peak.y,
      xend=chromEnd, yend=peak.y),
      color=model.color,
      size=1,
      data=peaks.dt)+
    geom_point(aes(
      chromEnd, peak.y),
      color=model.color,
      shape=21,
      fill="white",
      data=peaks.dt)+
    theme_bw()+
    theme(panel.spacing=grid::unit(0, "lines"))
}

## To analyze computational complexity, plot number of intervals
## stored in cost function, versus data point, for each cost status.
imat <- fit[["intervals_mat"]]
interval.dt <- data.table(
  intervals=as.integer(imat),
  status=c("peak", "background")[as.integer(col(imat))],
  data.i=as.integer(row(imat)))
if(require("ggplot2")){
  ggplot()+
    scale_fill_manual("label", values=ann.colors)+
    geom_rect(aes(
      xmin=firstRow-0.5, xmax=lastRow+0.5,
      ymin=-Inf, ymax=Inf,
      fill=annotation),
      alpha=0.5,
      color="grey",
      data=fit[["label_dt"]])+
    geom_line(aes(
      data.i, intervals, color=status),
      size=1,
      data=interval.dt)
}

Convert data for input to FLOPART

Description

FLOPART needs at most one label per coverage data row, which may not be the case for arbitrary coverage/labels.

Usage

FLOPART_data(coverage, label)

Arguments

Value

named list: coverage_dt is data table representing a run-length encoding of the input coverage data, with additional rows if there are label chromStart/chromEnd values not in the set of coverage positions; label_dt is a data table with one row per label, and additional columns firstRow/lastRow which refer to row numbers of coverage_dt, 0-based for passing to C++ code.

Author(s)

Toby Dylan Hocking

Examples

library(data.table)
d <- function(chromStart, chromEnd, count){
  data.table(chromStart, chromEnd, count)
}
(cov.dt <- rbind(
  d(0, 10, 53),
  d(10, 20, 124)))
l <- function(chromStart, chromEnd, annotation){
  data.table(chromStart, chromEnd, annotation)
}
lab.dt <- rbind(
  l(2, 7, "noPeaks"),
  l(8, 15, "peakStart"))
FLOPART::FLOPART_data(cov.dt)
FLOPART::FLOPART_data(cov.dt, lab.dt)
data("Mono27ac", package="FLOPART")
sapply(Mono27ac, dim)
converted <- with(Mono27ac, FLOPART::FLOPART_data(coverage, labels))
sapply(converted, dim)

Interface to FLOPART C++ code

Description

Interface to FLOPART C++ code

Usage

FLOPART_interface(
  data_vec,
  weight_vec,
  penalty,
  label_type_vec,
  label_start_vec,
  label_end_vec
)

Arguments

Value

List with named elements: cost_mat and intervals_mat (one row for each data point, first column up, second down), segments_df (one row for each segment in the optimal model)

H3K27ac ChIP-seq data from one Monocyte sample

Description

Raw coverage data and labels are provided in order to test the FLOPART algo.

Usage

data("Mono27ac")

Format

List of 2 data.tables, coverage and labels.

Source

https://github.com/tdhock/feature-learning-benchmark

Smaller H3K27ac ChIP-seq data from one Monocyte sample

Description

Raw coverage data and labels are provided in order to test the FLOPART algo.

Usage

data("Mono27ac.simple")

Format

List of two data tables, coverage and label.

Source

https://github.com/tdhock/feature-learning-benchmark

Lookup the integer values used to represent different label types

Description

Lookup the integer values used to represent different label types

Usage

get_label_code()

Value

Integer vector with names corresponding to supported label types