Release 0.2.x

Release 0.2.4 (17/12/2016)

  • fix some bugs
  • supporting gtf and gff3 and remove --anno_type
  • change default theta1 as 3, meaning more flat prior.

Release 0.2.3 (04/11/2016)

  • fix some bugs
  • change default save_sample into 0

Release 0.2.2 (29/10/2016)

  • fix some bugs
  • simplify the arguments
  • change default theta1 as 10
  • change default save_sample (original sample_num) into 500
  • keep transcript in diceseq results in the same order as in gtf file
  • To get the same default as v0.2.1, please set --mcmc 0 20000 1000 100 --thetas 3 None

Release 0.2.1 (16/06/2016)

  • updated calculating TOTAL_COUNTS for pysam 0.90
  • update documentary by adding more examples and tutorials

Release 0.2.0 (16/05/2016)

  • fixed some bugs
  • cleaned some codes

Release 0.1.x

Release 0.1.9 (29/01/2016)

  • fixed some bugs
  • tested the estimate of bias parameters and fragment length distribution
  • introduced the parallel computation, new you can run it on multiple cores
  • changed the format of output file for diceseq, which is easier to read
  • introducted the auto detection of using single-end or paired-end reads
  • removed the EM algorithm for buin-in, as we could easily use multiple cores to reduce running time

Release 0.1.7 (21/12/2015)

  • largely improved for the case with multiple (more than 10) isoforms, and introduced EM algorithm for buin-in
  • extended for python 3
  • improved bias estimates
  • not rely on h5py anymore, by design diceseq format for bias parameter and samples
  • introduced an out_utils.py for downstream analysis of diceseq results
  • fixed some bugs on paired-end reads matching and processing
  • removed reads simulation, as it can be done by spanki

Release 0.1.6 (10/11/2015)

  • slight change to be compatible with pysam 0.8.3
  • change numpy.math.erf() to personally wroten erf(), as the former is not supported in numpy 1.8.2 and higher
  • provide the option to not save the hdf5 copy
  • remove dice-static for direct running

Release 0.1.0 (26/10/2015)

  • probabilistic models
    • finished dynamic model: estimate jointly
  • dirrect running file
    • diceseq: support separated and joint estimates
    • dice-static: not suggested anymore.
  • bug fixed
    • reads simulation

Release 0.0.x

Release 0.0.1 (01/09/2015)

  • first version of DICEseq

  • utils objects or functions
    • FastaFile: read and write fasta file
    • BiasFile: read and write bias parameters
    • ReadSet: get reads location index of genome region
    • sam_utils: load and preprocess RNA-seq reads
    • gtf_utils: load gtf file as list of Gene and Transcript objects
    • TranUnits: processing at isoform level (or transcript level)
    • TranSplice: processing on multiple isoforms (or gene level)
  • probabilistic models
    • static model: estimate indivually
    • dynamic model: estimate jointly (under study)
  • dirrect running file
    • diceseq
    • dice-static
    • dice-count
    • dice-simulate
    • dice-bias
    • dice-biasplot