TO DO: o change optim parameter in main flipflop o check everything ok on windows and os o take care of all cpp warnings (isolasso+spams) CHANGES IN VERSION 1.6.1 -------------------------- o add the header ie the sample names to the output tables (created when using output.type='table') o correct little bug concerning the SAM file when header is present (see samio.cpp first test over linecount replace by totalnumread) CHANGES IN VERSION 1.5.16 -------------------------- o correct a bad memory managment in readgroup.cpp (do not use erase on a vector!) CHANGES IN VERSION 1.5.15 -------------------------- o correct a bug that may happen when the pre-processing encountered a * in the CIGAR field. now it simply ignore the read (align.cpp). CHANGES IN VERSION 1.5.14 -------------------------- o correct a bug which might happened when reading the RG:Z tag in the SAM file (align.cpp) o allow H characters in the CIGAR field o correction in spams (prox/project.h) related to topological order CHANGES IN VERSION 1.5.12 -------------------------- MAJOR CHANGES o parallelization -----> FlipFlop can be run on multiple cores. NEW FEATURES o 2 novel options related to the parallelization: - 'sliceCount' number of slices in which the pre-processing '.instance' file is cut. when OnlyPreprocess is set to TRUE, it creates slices and you can run FlipFlop independently afterwards on each one of the slice. - mc.cores it automatically distributes the slices on several cores when sliceCount>1. when using a given preprocessing.instance and several samples, it uses several cores for some operations concerning multiple samples. CHANGES IN VERSION 1.5.11 -------------------------- o added the fraction of each predicted transcript (among one gene) in the output GTF file. CHANGES IN VERSION 1.5.10 -------------------------- MAJOR CHANGES o implementation of a MULTI-SAMPLE version of FlipFlop -----> the multi-sample procedure uses several samples simultaneously in order to increase statistical power. by default the multi-sample procedure is a penalized likelihood with the group-lasso penalty. you can also choose to perform a simple refit after using the pooled data. NEW FEATURES o new features as a result of the multi-sample extension. Novel options are - 'samples.file' optional samples file (one line per sample name) - 'output.type' type of output when using several samples simultaneously. when equal to "gtf" the output corresponds to a gtf file per sample with specific abundances. when equal to "table" the output corresponds to a single gtf file storing the structure of the transcripts, and an associated table with the abundances of all samples. - 'mergerefit' if TRUE use a simple refit strategy instead of the group-lasso. Default FALSE. MINOR CHANGES o change a bit the paired-end graph paradigm. Ensure that one isoform is one unique path in the graph, as for single-end graph. o [parameter] add the delta (poisson offset in the loss) parameter as an option, in case of fine-tuning. o [parameter] the cutoff option is now active for both single and paired, and the default value is 1%. CHANGES IN VERSION 1.5.8 -------------------------- o change default MinCvgCut value from 0.25 to 0.05 CHANGES IN VERSION 1.5.7 -------------------------- o correct a stupid error (in flipflop.R) for calculating the total number of mapped fragments in the paired-end case CHANGES IN VERSION 1.5.6 -------------------------- MAJOR CHANGES o faster single-end and paired-end graph implementation o use_TSSPAS: if yes then start/end bins are the ones with no entering/outgoing junctions (+ annotated ones if given annotations) MINOR CHANGES o remove the sanity-check for possibly duplicated types, as it was solved into readgroup.cpp/getType o [pre-processing] back to default 1 for minJuncCount (I changed my mind !!) o [pre-processing] cancel change of version 1.3.9 (extreme boundary of segments calculateBound in readgroup.cpp) (I changed my mind !!) CHANGES IN VERSION 1.5.5 -------------------------- o allow more easy 2 steps run (preprocessing+prediction). The option "OnlyPreprocess" allows to perform preprocessing only. It creates two files 'prefix.instance' and 'prefix.totalnumread'. Running then flipflop with the option preprocess.instance='prefix.instance' performs the prediction (it reads automatically the total number of reads in prefix.totalnumread) o [pre-processing] push the default minJuncCount from 1 to 2 (high enough for a default?) o [pre-processing] slight change for paired-end preprocessing, see option to processsam (--single-only not automatic anymore) o start using Rprintf in c/cpp codes for user messages (to continue) CHANGES IN VERSION 1.5.4 -------------------------- o [pre-processing] solve small issue on CIGAR character 'S,X,=' (again!), see align.cpp CHANGES IN VERSION 1.5.3 -------------------------- NEW FEATURE o [pre-processing] add minJuncCount as an option, ie number of required reads to consider a junction/boundary as valid. CHANGES IN VERSION 1.5.1 -------------------------- o [pre-processing] solve small issue of possibly duplicated read type, see getType in readgroup.cpp CHANGES IN VERSION 1.3.9 -------------------------- o [pre-processing] redefine extreme boundary of segments (first and last), see calculateBound in readgroup.cpp CHANGES IN VERSION 1.3.7 -------------------------- o solve non-portable flags issues, see Makevars CHANGES IN VERSION 1.3.5 -------------------------- o solve compilation isue on Mavericks (using clang), see Makevars. CHANGES IN VERSION 1.3.4 -------------------------- o properly take into account 'S', 'X' and '=' characters (soft clipping, strict mismatch and match) in the CIGAR of the sam file CHANGES IN VERSION 1.3.2 -------------------------- o Solve small bugs in the graph construction o Update the optimization solver for refit. More accurate and faster! o Remove warnings from spams (to continue ...) CHANGES IN VERSION 1.1.6 -------------------------- NEW FEATURES o Add 'expected.counts' in the output objects. It gives the expected raw count (ie number of mapped fragments) for each predicted transcripts. This information is also available in the output gtf file. CHANGES IN VERSION 1.1.4 -------------------------- NEW FEATURES o Add a 'OnlyPreprocess' option for performing only the pre-processing step of the input sam file. This step writes two files: one file named 'prefix.instance' and one other named 'prefix.totalnumread', where 'prefix' is the prefix of the input sam file. The 'prefix.instance' file can then be given to the option 'preprocess.instance' and the total number of mapped reads is stored in the 'prefix.totalnumread' file. o Allow to give 'NN' (total number of mapped fragments) even when using the sam file as input. This can be used to run flipflop in parallel on parsed sam files with the same NN constant. USER-LEVEL CHANGES o Handle '~' in input paths with path.expand function. o Give a more detailed R output with the read count. MINOR CHANGES o Do not write individual Coverage into the pre-processing file anymore ie, the .instance file. (in practice comments lines in readgroup.cpp, part 'toStream') CHANGES IN VERSION 1.1.2 -------------------------- BUG FIXES o Switch from GRangeList to a regular list as the number of metadata columns could not vary in GRangeList.