3.2

@MatthewThe MatthewThe released this Feb 5, 2018 · 9 commits to master since this release

Assets 8
  • Changed FDR calculation for initial direction selection to the more liberal decoys/targets (#208)
  • Added parameter file option instead of only command line specification (#149)
  • Allow training only on best PSM per peak group or spectrum (#207)
  • Include spectral counts for lists of protein inferences (#202)
  • Allow separate control of FDR threshold for initial iteration (#190)
  • Added check of deltaMass and retentionTime columns with DOC features (#167)
  • Added check if output files are writable before training SVM (#185)
  • Fixed tandem BioML input check for tandem2pin (#180)
  • Fixed discrepancy between windows and linux results (#203)
  • Fixed fido protein grouping (#201)
  • Fixed zlib DLL issue on windows (#197)
  • Fixed/clarified several log and help messages (#199, #205, #179)

@MatthewThe MatthewThe released this Jul 31, 2017

Assets 3
  • Fixes empty output on xxx2pin converters on MacOS (#178).

NOTE: for all other platforms besides macOS, the 3.1 release below is still the latest release.

@MatthewThe MatthewThe released this Jul 27, 2017 · 1 commit to branch-3-01 since this release

Assets 3
  • Fixes macOS installer problems with macOS 10.11+ (#193)

NOTE: for all other platforms besides macOS, the 3.1 release below is still the latest release.

3.1

@MatthewThe MatthewThe released this Nov 17, 2016 · 41 commits to master since this release

Assets 8

v3.01

  • Changed all FDR calculations to the more conservative: (decoys+1)/targets
  • Added -o/-no-terminate flag to allow overriding of fatal errors:
    • No default direction can be found: set to the first feature (#135)
    • Separation between target and decoy PSMs/proteins is too great: continue with this direction anyway
    • No target PSMs present: return empty result list
    • No decoy PSMs present: give all target PSMs a q-value of 1/numPSMs
    • No features present: create a single pseudo feature which is always zero
    • Too many values in default direction vector: trim default direction vector to number of features
  • Added mix-max method for FDR and PEP calculation, which replaces target-decoy competition as the default method. The advantage is that this will report all input PSMs in the PSM output and also corrects for incorrect native spectra in qvality (#152)
  • Automatic detection of search input as coming from a concatenated search or from separate target and decoy searches. For a concatenated search, the mix-max method will be skipped. For separate target and decoy searches the user can choose to use the mix-max method (default) or target-decoy competition (-Y flag)
  • Updated and improved stderr output to be more informative
  • Fixed fasta in-silico digestion for protein grouping which mistook N-methionine cleavage for a non-enzymatic cleavage site (#148)
  • Fixed retention time NaN predictions by setting them to 0 instead (#150)
  • Fixed protein group representation for picked-protein in the case that protein names contain a comma, by replacing these by semicolons (#153)
  • Fixed tandem2pin issue that labeled all PSMs as decoy when providing separate target and decoy input files (#154)
  • Added OpenMP runtime library to be included in windows installer (#158)
  • Fixed msgf2pin converter issue with C-terminal modifications (#161)
  • Fixed tandem2pin to parse PTMs (#132)
  • Fixed install location on windows 64-bit from "Program Files (x86)" to "Program Files" (#144)
  • Several fixes for OSX build script (#159, #162, #165)
  • Fixed mass difference calculation in certain converters (#166)
  • Fixed several build issues on several platforms (#79, #113, #133)
  • Build scripts are now changed and tested under CentOS 7.2 and Fedora 23, the precompiled binaries below are generated on those systems as well.
3.0

@MatthewThe MatthewThe released this Jun 9, 2016 · 72 commits to master since this release

Assets 8
  • Added protein-level FDRs using best-scoring peptide approach
  • Added support for training on subsets of PSMs to speed up analysis for large data sets
  • Fixed msgf2pin issue with huge ScoreRatio feature (#142)
  • Fixed tandem2pin issue with incorrect decoy label assigment for shared peptides (#143)
  • Added error message when no decoys are provided (#145)
  • Fixed reporting of XML parsing errors for converters (#146)

@MatthewThe MatthewThe released this Jul 7, 2015 · 102 commits to master since this release

Assets 7
  • Made Target-Decoy competition the default search mode (can be switched off with the new '-y' flag)
  • Fixed unbalanced fold sizes in cross validation
  • Fixed build script for Elude on Windows (#106)
  • Fixed enzC feature in the presence of a modification on the C terminal (#118)
  • Resolved duplicate -W flag (#124)
  • Fixed issue with Target-Decoy competition (#128)
  • Added support for modifications of the form P[mod_string]EPT of sqt input file for sqt2pin (#131)
  • Simplified the Fido interface for protein inference
    • Removed options to turn off optimizations that did not alter the posteriors
    • Proteins are always grouped if they have identical evidence and get a protein group number in the protein tab delimited output

@MatthewThe MatthewThe released this Apr 3, 2015 · 113 commits to master since this release

Assets 7
  • Added CentOS building script, tested on a clean install of CentOS5.9 (#98, #99)

  • Added support for reading from stdin using an empty hyphen argument (#107)

  • Added error message when running converters without arguments (#108)

  • Fixed invalid pin output from converters with --aa-freq parameter (#110)

  • Added exp_mass and calc_mass to the tab delimited pin format, so that it appears in the xml output (#112)

  • Fixed backwards compatibility with old MSGF+ output (#115)

  • Fixed sqt2pin issue of skipping second modification if there are two consecutive PTMs (#116)

  • Added OpenMP support for a maximum of 3 cores (1 for each of the cross validation bins).

    • runs about 2 times as fast, instead of the theoretical 3 (#119).
    • memory usage is slightly higher (<5%) than the single core implementation, since some structures had to be triplicated
    • the number of cores can be set to 1 dynamically by setting the environment variables
    export OMP_DYNAMIC=FALSE
    export OMP_NUM_THREADS=1
    
Jun 22, 2014
Added xml tags
Jun 22, 2014
Added xml tags