Next: Acknowledgment Up: Algorithms and Thermodynamics for Previous: EXAMPLE 3

Future plans

The Unix version of mfold will remain. The newest programs, such as plt22ps and plt22gif are written in command line mode. The older programs have shell scripts or Perl ``wrappers'' around them to make them appear as single binaries that operate in command line mode. The trend will be to replace older code as necessary with non-interactive programs. This makes it easier to piece together different programs to create new forms of output.

When mfold was first created, the limitations of personal computers did not make a PC version practical. This has changed radically in the past 10 years, and an Intel/Windows PC is now a fine environment for running mfold . The basic Fortran and C programs have already been ported and will run under Widows, but a Unix-like shell is necessary. The RNAstructure program is now a faithful recreation of mfold in Windows, with a convenient user interface. The major problem with the existing setup is that the Unix and Windows versions are totally different and will have to be updated in parallel to keep them equivalent.

The mfold programs running on SGI/Irix, SPARC/Solaris and Intel/Solaris have been incorporated into a world wide web (WWW) server that allows users from around the world to submit sequences for folding. This server has some extra features not available in the mfold package, and goes well beyond the very simple HTML output of the current mfold software. However, this software offers nothing new in terms of predictions, and it will be described elsewhere. Rapid developments in web browsers and languages such as Java-script and Java may make an HTML (or similar) interface to mfold better than others. As things stand now, the mfold server can run on a (local) Unix computer and be accessed by web browsers running on personal computers.

Additional parameters will be added to the command line version of mfold in the future. These will be described when the mfold command is given without parameters and the documentation will be altered accordingly. In the near future, a base labeling frequency will be added so that the user can specify the frequency of base enumeration. Other controls on secondary structure, such as zooming on images about specified coordinates, could be added, but these are already available through the use of the plt22ps and plt22gif programs.

The ``energy'' parameters from the older, interactive versions of mfold could easily be introduced in command line form. This would make it unnecessary to run the nafold program directly to alter them. Table 5 lists these parameters that are not defined in the MISCLOOP file.

Current plans call for the addition of coaxial stacking to the folding algorithm and possibly the creation of a special version that uses Jacobson-Stockmeyer theory to assign more realistic free energies to multi-branch loops, as in equation 7. In addition, a practical approach to pseudoknots will be attempted, where pairs of mutually exclusive helices that create pseudoknots are identified in the energy dot plot . In these case, the bases involved in 1 or perhaps 2 of these pseudoknots can be constrained to be single stranded, and foldings predicted to fill in the rest of the secondary structure.

Another future development will be the introduction of a 2 molecule folding system. This immediately complicates the problem, since concentration now becomes important. In addition, the folding of certain very simple bi-molecular systems is at least as hard as predicting pseudoknots in the folding of a single sequence.

---

Next: Acknowledgment Up: Algorithms and Thermodynamics for Previous: EXAMPLE 3

Michael Zuker Center for Computational Biology Washington University in St. Louis 1998-12-05