NEWS

---

epigraHMM 1.9.1

• Fix normalization of log-probabilities in cpp code to avoid underflow

epigraHMM 1.7.2

• Fix callPeaks writers so that it works on Windows

epigraHMM 1.7.1

• Fix callPeaks function to export .bed, .wig, and .bedGraph files
• Output files are now overwritten, if existing

epigraHMM 1.3.2

• Exporting function estimateTransitionProb to estimate transition probabilities from a sequence of states of a Markov chain

epigraHMM 1.2.1

• Bug fix of output paths to handle paths with '.'

epigraHMM 1.2.0

• Second release of epigraHMM on Bioconductor.

epigraHMM 1.0.8

• Minor bug fix in maxStepProb documentations

epigraHMM 1.0.7

• Exporting maxStepProb, which compute the MLE of initial and transition probabilities of a K-state HMM, as well as simulateMarkovChain, which simulates a Markov chain of length 'n' given a matrix of transition probabilities

epigraHMM 1.0.6

• Minor bug fix in controlEM documentation

epigraHMM 1.0.5

• Minor bug fix in callPatterns and info function (explict import of S4Vectors::mcols and utils::tail).

• Exporting expStep function, which implements the E-step of EM algorithm (forward-backward & Viterbi algorithm) for a K-state HMM.

epigraHMM 1.0.4

• Adding function callPatterns to exp[ort] combinatorial patterns (or posterior probabilities) associated with a given set of genomic regions.

• Adding function info to print summary statistics from epigraHMM output. This function will print the model's BIC, log-likelihood, and combinatorial patterns associated with mixture model components.

• Adding new example dataset helas3 with ENCODE ChIP-seq data from broad epigenomic marks H3K27me3, H3K36me3, and EZH2.

• Adding option to prune combinatorial patterns associated with rare states. See vignette for details.

• In differential peak calling, epigraHMM now exports combinatorial pattern table. See vignette for details.

• Improvement of the vignette to clarify epigraHMM's use of blacklisted regions and gap tracks.

epigraHMM 1.0.3

• Minor updates in the NEWS file as well as the README page.

epigraHMM 1.0.2

• epigraHMM now exports a function called segmentGenome that segments a given genome (e.g. 'mm10') into non-overlapping genomic windows while considering gap tracks and blacklisted regions.

epigraHMM 1.0.1

• Minor fix in the package DESCRIPTION file and version numbers

epigraHMM 1.0.0

• First release of epigraHMM on Bioconductor.

• It is now possible to add normalizing offsets via addOffsets.

• epigraHMM now uses hdf5 files to store all intermediate data during computation of the EM algorithm. Intermediate data include window-based HMM and mixture model posterior probabilities, and forward-backward probabilities. This change leads to a better memory utilization of the package.