Release

Paul Deveau

2017-03-02

• Added normalization by contamination if “keep.all.model”
• Added clause to check if normalized.center exist in evolution plot

Release of 2/10/2016

• Added corrective term for weight of each cluster;
• Switched back default model selection to “BIC”.

Release of 29/9/2016

• Changed optimization procedures (see One_step_clustering documentation);
• Added grzero: returns value that zeros the gradient (used in diploid cases);
• Added tests.

Release of 27/9/2016

• Fixed gradient computation;
• Removed alpha (to force possibility)
• Switched from computation of dbinom to log(dbinom) by default;
• Added testthat : plots, QuantumCat,Flash for BIC and overdiploid;
• Fixed QuantumCat with single sample

Release of 23/9/2016

• Added computation of gradient for “optim” and “optimx”;
• Corrected priors creation from FLASH for QuantumClone;
• Added jitter if number of initializations is >1 and preclustering with FLASH.

Release of 22/9/2016

• Fixed errors from FlashQC;
• Updated validation tools.

Release of 20/9/2016

• Added integration of binomial density over 2xEpsilon interval;
• Added computation of Precision, Recall, F1, Rand Index in Precision_Recall;
• Changed maxit to Initializations.

Release of 12/9/2016

• Changed parameters for DEoptim controls;
• Solved tree generation issue if the progeny can only come from one node.

Release of 12/9/2016

• Solved fik*log(fik) when fik = 0 issue in FlashQC;
• Added controls to DEoptim;
• Prepared for integration of RcppDE.

Release of 12/9/2016

• Changes in weight computation: normalization by sum of weights;
• Added DEoptim solver for EM;
• Flash QuantumClone handles polyploid regions;
• Flash QuantumClone handles BIC, AIC or numeric correction of BIC for cluster selection.

Release of 9/9/2016

MAY BE UNSTABLE

• Added FlashQC : using hierarchical clustering and distance on the p-value of 2 variants being from the same distribution;
• Updated parallel: initialization and number of clusters are run in parallel, not independently;
• For rearranged genomes, after the first clustering, a second one is done after removing unlikely states;
• Preclustering can now be done with FlashQC

Release of 31/08/2016

• Rewrote this vignette in html
• Added reproducible testing for PyClone, sciClone and k-medoid
• Updated figure dpi for vignettes

Release of 19/08/2016:

• Corrected NMI computation if a number of cluster is missing (for example cluster values can be 1:3, 5:7, cluster value 6 would be considered missing);
• Changed the proof of concept pipeline to be easier to read and edit;
• Changes in Probability.to.belong.to.clone: now returns filtered and unfiltered probabilities (clustering is on filtered probabilities).

Release of 16/08/2016:

• Corrected hard clustering parameters;
• Corrected Probability.to.belong.to.clone;
• Gain from parallelization script.

Release of 12/08/2016:

• Computed PoC for clustering quality;
• Changes in eval.fik for an expected 5-10% gain in computational speed.

Release of 9/08/2016:

• Removed 3D plots to prepare CRAN release

Release of 8/08/2016:

• Removed rgl dependency (3D plots), causing QuantumClone to be unavailable on Ubuntu;
• Updated pipeline tests;
• Computational time test compiled;
• Ongoing:
• Issue in the distance error (position of the cluster for drivers vs real position);
• Compile Rmd for computational clustering quality results.

Release of 4/08/2016:

• Added function to easily compute Normalized Mutual Information from QuantumClone output - for simulated data;
• Added pipelines to compare to paper pipeline.

Release of 29/07/2016:

• Option to force single copy status - i.e. integrate QuantumCloneSingle in QuantumClone
• Perspectives:
• tidy plot code in cluster_plot;
• Reproducible test in R (QC only);
• Reproducible testing in R+bash (QC+sciclone+pyClone)

Release of 28/07/2016:

• Added AIC (Akaike Information Criterion);
• Added modified BIC (user input);
• Added possibility to use optimx instead of optim for the maximization step;
• Using compiler package to improve computing time;
• Improved code readability;
• Improved parallel computation handling;
• Solved conflicting name in simulation: cellularity of clone is now refered to as cellularit;
• Added option to keep all iteration models;

Release of 10/12/2015:

• Changed graph output to pdf;
• Fixed non-clustered graph name file

Release of 12/11/2015:

• Added release tag on github;
• Checked for ggplot2 scheduled release

Release of 2/11/2015:

• Correction of evolution plot;
• Adding content to vignettes;
• Added phylogenetic tree reconstruction and representation, along with examples;
• Added ImportFrom according to notes from R-devel: http://win-builder.r-project.org/;

Release of 30/10/2015:

Major fix

• Post processing affected order of mutations but not of clustering. The clustering thus appeared erroneous in version 0.9.15 and 0.10.15;

Minor changes

• Adding vignette release;
• Changed dependency of doSNOW to doParallel;
• Added data to decrease loading and check time;
• Added function to plot results from output (by 2 elements);
• Added function to plot cellularity of clone in each sample (for time series, with proportion of mutation in each clone);