Center for Integrative Bioinformatics Vienna
Max F. Perutz Laboratories
Dr. Bohr Gasse 9
A-1030 Vienna, Austria

Older versions:

IQPNNI 3.3.1 (built 8th December 2008):

  • Non-parametric bootstrap
  • Bug fixes

IQPNNI 3.1 (built 13 Mar 2007):

  • Codon model: The program goes through two stages. At first the tree is reconstructed based on HKY model for DNA. Then it applies one of the following codon models for inference of positively selected sites:
    • NY98 (Nielsen & Yang 98): main model for inferring positive selection.
    • YN98 (Yang & Nielsen 98): special case of NY98 with 1 Ns/Sy category.
    • GY98 (Goldman & Yang 94).
    • CP98 (Pedersen et al. 98): model incorporating CpG depression.
    • CGTR: GTR-analog for codon models (unpublished).
  • Gamma + Invariable sites rate heterogeneity.
  • Site-specific rates (Meyer & von Haeseler 2003) improved. Also write out site-rates based on empirical bayesian if gamma rate is specified.
  • New protein models: rtREV (Dimmic et al. 2001), user-defined model by a file containing amino-acid replacement rates and frequencies.
  • Warning if number of iterations is too small as recommended by the stopping rule.
  • New command line options.
  • Bugs fixed:
    • Zero state frequencies: they are now replaced by a very small number.
    • Checkpoint: now correctly recovered from stopped point.
    • Restriction on number of sites: from limit 100,000 to unlimited now.
  • Bugs identified:
    • Parallel version on Infiniband system under MPICH.

IQPNNI 3.0.1 (built 4 Sept 2006):

  • Zero iteration: if user specifies number of iterations to be zero, the program will only evaluate the starting tree (either BIONJ or user-defined tree) by optimizing model paramters and branch lengths.
  • Triplet tree: the program can now run on alignment of just 3 sequences.
  • Scaling technique to avoid numerical underflow on large datasets. It now can stably analyze alignments with more than 1,000 sequences.
  • At least twice faster than v3.0. The "long double" datatype is replaced by "double", making it more compatible to most computers.
  • Memory consumption is reduced at least by half by a new mechanism of storing conditional likelihood vector.
  • New eigensystem adapted to reversible instantaneous rate matrix.


  • The program runs at least twice faster than version 2.x by applying Newton-Raphson method to optimize branch lengths of the tree and some other algorithmic means.
  • Run in parallel with the Message Passing Interface (MPI) library, which can be obtained from e.g. Available MPI Implementations.


  • General time reversible model of evolution.
  • Site-specific subsitution rates: Sonja Meyer and Arndt von Haeseler, Identifying Site-Specific Substitution Rates, Mol. Biol. Evol. 20(2):182-189.2003.
  • Check pointing: If the program was crashed or stopped by users, it can continue from the last stopped point.


Version 3.1

Version 3.0.1

Version 3.0

(BUG FIXED: 20th Dec 2005)
(PDF) NEW IN 3.0 (Please read carefully, especially when you change from 2.6)

Version 2.6

  • Please, let us know if you download this program by sending an email to {minh.bui,arndt.von.haeseler}@univie.ac.at

Printable version of: http://www.cibiv.at/software/iqpnni/oldversions.html

contact imprint .