Once the trial conformations and experimental data was entered it is recommended to save the project through DISCON -> Write DISCON project.

This saves all of the backcalculated values to a JSON file where if it is necessary can be visualized by third party software (eg http://jsonviewer.codeplex.com/).

To perform a DISCON calculation select "Run DISCON Job" under the DISCON menu. The software will ask you to give a name for the calculation. As repetitive calculations

are necessary it is recommended to save the job under a new folder.
 

1. Cluster Level

The software will require the number conformations needs to be employed for the fitting. The notation gives access to run consecutive clustering level calculations at the same time.

If 5 is entered as the number of conformations, DISCON will cluster all of the conformations into 5 families where clustering is performed according to the backcalculated NMR variables.

This strategy enables quickly differentiating the conformations according to the measured experimental values.

2. Parameters

NOE and Couple Weight : The weight of NOE and  J values needs to be adjusted to be able to employed together. If NOEs are given as volumes they are normalized to 1 where the J values vary between 0-10 Hz. It is recommended to give a 10:1 ratio for the fitting but the user can provide different ratios to see how well the data are fitting.

Use of Parameters in Calculations: User can select to use only J or NOE values for validation purposes. A more important option is to use selected parameters for the calculations which is quite useful if a variable is found to be erroronous. To activate selected parameters select the "Use selected parameters" option and left click with mouse on the NOE and J Tables to select the variables. Multiple selections can be performed by pressing down  the Ctrl button in PC and Command Key in Mac. When Shift key is pressed consecutive selections can be done.

The selected parameters will be utilized in the calculations where the rest of the parameters will be ignored.

3. Conformations

Selected conformations can be used for the calculations. If this option is selected all the non utilized conformers will be clustered into a cluster and their fitting value will be set to 0.

The cluster settings are very important for getting reliable results. If the trial conformers were given in an order of increasing energy the representative members of the clusters that will be selected will correspond to the lowest energy member of that cluster family if "Use smallest members of the clusters" option was selected.

The second option is to use random members of the cluster and repeat the fitting studies n times with different cluster members. It is recommended to run as many trials as possible to see if the distribution is uniform with respect to the members of the cluster families. Our studies show that for clustering level 5 of 1000 conformers may require 1000 repetitions for getting the best results.

The last option will cause to use all conformations in the fitting, a scenario suitable for a limited number of trial structures. If this option was selected for a large set of trial conformers it will generate an even distribution.

4. Genetic Algorithm Parameters for Global Optimization Algorithm

The most important parameters are the population size and the number of generations. A typical run for 10 experimental variables and 50 conformations a population size of 5000 and 1000 number of generations is recommended. GA Mode can be altered to 1 to 4 for different genetic algorithm approaches.

DISCON calculations generally take 5 to 10 hours for small size molecules where multiple runs and examination of the results at multiple cluster levels are required. It is the users' responsibility to check the statistical significance of the results and the convergence of the data. Multiple selection criteria as well as limiting conformations and experimental parameters are necessary to find the outliers and the important conformations. Examination of repetitive results will reveal the right cluster level and conformational families.