-----------------------
Release 1.0.10
-----------------------

The following minor changes have been made:

  1. Starting values no longer need to satisfy bound constraints or linear frequency constraints. This increases the chance of finding viable starting values.
  2. The references have been updated to point to the SAGMB paper.

-----------------------
Release 1.0.9
-----------------------

The stand-alone C code and the web application now also accept MEME-style background files.

-----------------------
Release 1.0.8
-----------------------

The following changes and additions have been made in the R package:

  1. cosmo() now accepts a file for specifying the background Markov model that follows the MEME standard.
  2. The new function bfile2tmat() converts a MEME-style background file into a transition matrix.
  3. The new function bgModel() produces an estimate of the transition matrix for the background Markov model.
  4. A useful error message is printed if input sequence file or background sequence file contains a sequence of length zero.
  5. Default options for the case that data set contains a single sequence have been updated.

-----------------------
Release 1.0.7
-----------------------

A number of default options have been changed on the basis of new simulation results:

  1. Default value for 'wCrit' changed from 'eval' to 'bic'.
  2. Default value for 'modCrit' changed from 'eval' to 'lik'.
  3. Default value for 'conCrit' changed from 'pwmCV' to 'likCV'.
  4. Default value for 'conTrunc' changed from '100' to '90'.

-----------------------
Release 1.0.6
-----------------------

Several changes and bug fixes:

  1. GUI for interactively constructing constraint sets has been moved into separate package 'cosmoGUI'.
  2. Functions for plotting DNA sequence logos have been moved into separate package 'seqLogo'.
  3. Added 'maxIntensity' parameter to cosmo() to allow likelihood to be maximized wrt to intensity parameter as well.
  4. Default value for 'intCrit' changed from 'eval' to 'lik'.
  5. simScore() now takes argument 'minOverlap' specifying minimal overlap between predicted and true motifs to declare a hit.
  6. Relaxed checks on starting values for satisfying constraints to avoid not finding any starting values.
  7. Uninitialized starting values are now skipped.
  8. Bug in computing E-value for sequences with unknown characters fixed.

-----------------------
Release 1.0.5
-----------------------

One minor bug fix in preparing C output for R.

-----------------------
Release 1.0.4
-----------------------

Two minor fixes:

  1. Calls to round() were replaced by the macro NINT() to make the code more portable.
  2. Problems in the computation of the AIC and BIC criteria were fixed.

-----------------------
Release 1.0.3
-----------------------

A few minor changes in the S4 classes used:

  1. The class constraintGroup has been removed. Simple lists of constraintSet objects are used in its place.
  2. The class postProbs has been removed. In its place, the slot probs in objects of class cosmo now contains a list with each element giving the posterior probability of motif occurrences along a particular sequence. Posterior probabilities are given a negative sign if the motif is more likely to occur in the reverse complement orientation than in the forward strand orientation.
  3. The class cosmo now contains a slot seqs that consists of a list of the sequences in the input dataset.
  4. The slots conGrp and roc.motifs in class cosmo have been removed.
  5. Posterior probability plots are now generated using the plot() method on objects of class cosmo with the argument type="prob", rather than by calling the plot() method on an object of class postProbs.
  6. A few other slots in the cosmo object have been renamed: cand.orders -> back, cand.model -> cand, transMats ->tmat, sel.model -> sel.

-----------------------
Release 1.0.2
-----------------------

A few minor changes and bug fixes:

  1. simScore() now calculates the area under the ROC curve as well. It now takes as input an align object and a cosmo object.
  2. rseq() can now simulate sequences of variable lenghts.
  3. The cosmo class now includes a slot of class align that ranks all potential start sites by posterior probability.
  4. The readFASTA() function (not exported) now closes any connection that was opened during the call.
  5. The example file seq.fasta now contains 10 rather than 20 sequences to make the vignette less bulky.
  6. The package now uses a CITATION file instead of the citation.cosmo() function.
  7. Global variables created by the constraint GUI are removed on exit.
  8. A minor bug in the selection of the order of the background model was fixed. The bug occurred with some datasets that contained unknown characters like 'N' or 'X'.

-----------------------
Release 1.0.1
-----------------------

A few minor changes and bug fixes:

  1. If the datasets contains only one sequence, only the TCM model will be considered. A 0th order Markov model is used for the background distribution. All other model parameters are selected based on the E-value by default.
  2. The TCM model does not report overlapping motifs anymore.
  3. A minor bug in creating the sel.model output slot was fixed.

-----------------------
Release 1.0.0
-----------------------

Initial release.