News & Updates

June 2020

• IMPORTANT ANNOUNCEMENT: Over the past year a serious matter has come to light regarding the validity of the GROMOS family of force fields and the GROMACS simulation engine. Recent benchmarking/validation evidence (see here and here for more) has shown that GROMOS, as implemented in GROMACS, may produce incorrect physical properties (e.g. density) for some solvents and lipids (although, curiously, not for proteins) with modern versions of GROMACS. This is due to the fact that GROMOS parameters have been developed using a computationally efficient, but physically inaccurate multiple time-stepping scheme, a charge groups model and a twin-range cut-off implementation, which are incompatible with the use of proper physical models, including the standard single-range Verlet cut-off used by most all-atom force fields, as well as the Trotter decomposition method of the MD integration algorithm implemented in GROMACS and all other modern MD simulation codes for more than a decade. Although GROMACS did support charge groups and a Trotter-compatible twin range cut-off implementation, these have been removed starting from version 2016.2 and onwards (and were deprecated in versions starting from 5.1.4 and leading up to 2016.2). However, the discrepancies of GROMOS with the single-range Verlet cut-off were only recently discovered.
  
  The above has led to a controversial situation, with GROMOS force field developers essentially "blaming" GROMACS for abandoning the twin-range cut-off, while the GROMACS developers themselves point out that the reported inaccuracies stem from the force field developers following a physically inaccurate parametrization scheme and using the obsolete 4.0.7 version of GROMACS (see the relevant publication here). As a result of this controversy, the use of GROMOS has been deprecated in GROMACS, with gmx grompp producing warnings regarding the force field's accurracy (see the GROMACS RedMine report for this here).
  Although the force fields themselves can still be used (by suppressing the warnings with the -maxwarn option), their use is highly discouraged. As of the most recent version (2020.2), this situation has not changed; furthermore, it has been suggested that unless the GROMOS force field is re-validated (and perhaps even re-parametrized) with modern simulation schemes, support for it may be completely removed in future GROMACS versions.
  
  After much deliberation, the GNOMM team has decided NOT to abandon support for GROMOS, especially because this force field has been proven to be capable in modeling the dynamics of β-sheet structures and β-barrels properly. Instead, we suggest that users interested in this particular force field use the GROMOS structures generated by GNOMM with GROMACS versions up to 5.1.4, which still support the group cut-off scheme and twin-range cut-offs. We will continue to closely monitor the situation and the progress of GROMOS validation, as well as the responses of GROMACS to the latter.

December 2019

• GNOMM has been transferred to a new web server, available from the default url (http://bioinformatics.biol.uoa.gr/GNOMM) as well as the alternative link http://thalis.biol.uoa.gr/GNOMM. The previous website is still available, for those users that want to obtain their results, but it will be decomissioned soon. Make sure to update your bookmarks!
  The new web server is faster and more powerful compared to the original one, meaning that submitted jobs also run faster. This is true for all available forces and especially for the time-consuming CHARMM36 option, which now requires about 1/5th of the time taken in the original server.

May 2019

• Created this "News & Updates" page.
• Added the Pseudomonas aeruginosa PAO1 LPS for the Martini and PACE force fields. The PAO1 LPS has been modeled using the topology proposed by Ma et al, J. Phys. Chem. B 119 (46): 14668-14682, 2015.
• The GNOMM paper has been assigned to issue Vol. 40 No. 18 (July 5th) of the Journal of Computational Chemistry! Update your EndNote references, if you haven't already.

April 2019

• During the March 25 - April 5 period, automatic structure retrieval from the OPM database was broken, due to the latter's changes in their PDB file organization system. This has been fixed and, from now on, automatic retrieval from OPM is once again available, albeit rather slow (due to OPM using Google Cloud for storage, which adopts authentification protocols that hamper programmatic retrieval schemes).

March 2019

• The GNOMM paper is available online! Read it HERE.