v2.0-bench

Benchmark files for the future 2.0 release in Ymir's format:
benchmark-alpha-full.tgz - TRA
benchmark-beta-full.tgz - TRB
Each archive have 4 files with 100,000, 250,000, 500,000 and 1,000,000 clonotypes.

v1.0

MAAG

1.0 version

• Implement MAAG (Multi-Alignment Assembly Graph).
  • Implement full probability computing.
  • Implement slice probability computing.
  • Add a fix for joint VJ prob computing.
  • Add copy constructors for MAAG and MMC.
  • Implement forward-backward algorithm for VJ receptors.
    • Implement joint nucleotide distribution computing VJ receptors.
  • Implement forward-backward algorithm for VDJ receptors.
    • Implement joint nucleotide distribution computing VDJ receptors.
• Implement MAAG builder.
  • Figure out MAAG builder interface.
  • Implement VJ MAAG builder.
  • Implement VDJ MAAG builder.
  • Optimise sizes of D deletions matrices.
  • Add a fix for joint VJ choosing.
  • Implement replaceEventProbabilities() function used for speed up statistical inference in models.
  • Pointers or copies in constructor?
    • Copies.
  • Figure out how to properly work with string iterators in insertions.
  • Check for 0-based or 1-based. All sequence building and other stuff should be 1-based, because it's the number of aligned characters.
  • Check edge cases, e.g., sequence positions vectors when all gene segment was deleted and insertions are inserted.

PAM

1.0 version

• Change name from ASM to something like Probabilistic Assembling Model.
• Add to ModelParameterVector joint VJ probabilities.
• Refactor ModelParameterVector for easier access to event probabilities.
• Implement basic PAM (probabilistic assembling model).
  • Implement model parameters reading from file procedure.
    • Add creating ModelParameterVector subroutine.
    • Add prob vector grow and shrink subroutines.
  • Implement computing probabiltiies procedure.
  • Implement clonotype assembler.
    • VJ
    • VDJ
  • Implement probabilities computing with recomputing the gene usage.
    • Implement
    • Test
  • Implement saving PAM's parameters to a file.
• Implement Murugan EM algorithm.
• Test EM algorithm and generation: generate an artificial repertoire and infer a model from it.
  • VJ
  • VDJ
• Add test for D gene segment length in parsing prob data.
• Add model parameter vector nulling procedure.
• Think about the proper way to parse inputs.
  • Always default input for Ymir, external python script for parsing various formats.
    • tcR data frames.
    • MiTCR data frames

Other

1.0 version

• Rename clones to clonotypes. (:
• Add P-nucleotide making.
• Add naive alignment step in the parser.
• Implement a naive local aligner for Diversity genes.
• Set up travis.
• Add error messages to parser if some problems with gene segments has been found.
• Add to parsers skip D alignment if genes are VJ.
• Add skipping of empty lines at the ends of files to all parsers.
• Add working drafts for default subroutines:
  • Computing.
    • C++ part.
    • Python 3 wrapper.
  • Inference.
    • C++ part.
    • Python 3 wrapper.
  • Generation.
    • C++ part.
    • Python 3 wrapper.
• Check for 1-based option in converters;

v1.0-pre3

First fully working pre-release version of Ymir. Download Source_code_FULL if you don't want to install one additional C++ library and download data for benchmarks and examples. See README for more information.

v1.0-pre2

update readme a bit; few minor changes

v1.0-pre

To benchmark unzip the benchmark.zip archive and put archived .txt files to $YMIR_PATH/benchmark/data/