tests

PeakSeg is a constrained maximum Poisson likelihood segmentation model described in ”PeakSeg: constrained optimal segmentation and supervised penalty learning for peak detection in count data” (source). We proposed a constrained Dynamic Programming Algorithm (cDPA) for computing a model that satisfies the PeakSeg constraints.

Installation

if(!require(devtools))install.packages("devtools")
devtools::install_github("tdhock/PeakSegDP")

Usage

There are two main functions for computing the constrained segmentation model:

• cDPA is a low-level interface to the C solver. Its inputs are N weighted data points and S, the maximum number of segments. It outputs a list with components named loss, ends, and mean (S x N matrices describing the solution).
• PeakSegDP is a more user-friendly wrapper of the cDPA. Its input parameter is P, the maximum number of peaks, which implies S = P*2+1. Its input data type is a data.frame with columns count, chromStart, chromEnd. It outputs a list of data.frames, peaks, error, segments, breaks.

Related work

• As explained in our ICML paper, the cDPA is a quadratic time algorithm that is not guaranteed to find the global optimum. For a linear time algorithm that recovers the global optimum, use the coseg package.
• For supervised peak detection in ChIP-seq data sets with several samples, see our newer method, PeakSegJoint.