Conformational generation requires 3D coordinates. These will be generated automatically by THINK if they have not been read in from a molecule file.
The conformational analysis implementation automatically detects rotatable bonds for the following classes of bonds:
and considers the specified number of positions (points) about each class of bond by incrementing the torsion angle about the bond. If zero positions are specified then all bonds of that class are ignored. The number of points about each class of bonds can be altered though the CUSTOMISE command.
THINK automatically applies an offset of 50% of the increment to each alpha bond during a systematic search. For instance, if the increment for alpha bonds is 60°, each alpha bond will be rotated by 30°, 90°, 150°, 210°, 270° and 330°. For conjugated bonds in sterically hindered environments, THINK will apply an offset of 50% of the increment before starting to rotate the bond thus sampling 90° and -90°. Crowded bonds are not sufficiently hindered conjugated bonds to ensure that they are non-planer. Sampling these bonds at 0°, 90°, 180° and -90° is often best.
The value for ring bonds is the maximum number of ring conformations to be considered, not the number of points for each ring bond. The torsion angles for ring conformations are stored in a lookup table. If the number of ring conformers required is less than the number stored then the lowest energy conformers are used.
During 3D and site searches (see sections 8.4 and 8.5 respectively), it may be desirable to limit the number of conformations generated for each molecule, and to ignore extremely flexible molecules, to prevent the search taking too long to complete. The common variable RCNFMX, when given a non-zero value, sets the upper limit for the number of conformations (a value of zero means that there is no limit). When a molecule is first scanned to detect the rotatable bonds, THINK automatically computes the total number of conformations that will be generated. If this exceeds the value of RCNFMX, each single (sp³-sp³) or alpha bond that has 12 points is reduced to 6 points until the total falls below RCNFMX. If necessary, the process is repeated for the single bonds, reducing the number of points to 3. If this still fails to bring the total below RCNFMX, the molecule is ignored.
The systematic search option sequentially increments the torsion angles about each rotatable bond in turn. This can be very time-consuming for flexible molecules, because it is comprehensive, but may be accelerated using the contacts check (see below). In addition, the conformational search may be abandoned after a certain time interval, specified through the CUSTOMISE TIMEOUT keyword. This is particularly useful in 3D and site searches (see sections 8.4 and 8.5) to prevent THINK spending a long time searching a single molecule.
Random sampling makes an estimate of the total number of conformations (computed by the product of the number of increments about each bond) and then uses a random number generator to select conformations randomly from within the estimated total. The implementation does not prohibit the same conformation being selected multiple times, and this will always occur if more conformations are requested than the total number of different conformers.
Random sampling does not necessarily produce the most representative set of conformers. A further sampling option is available, which attempts to space the sample conformations uniformly across conformational space. If conformations are rejected (for instance because of VdW contacts), then up to 10 attempts are permitted in order to find a sample interval which will explore additional conformations. The implementation ensures that the same conformation is not repeatedly sampled. However, this does not prohibit identical conformations to be output resulting from symmetry.
In order to prevent high energy conformers being considered, it is necessary to use the VdW or CPK contacts check. For systematic search, this also significantly reduces the time taken because when atoms in contact are detected, rotations about bonds which do not change the inter-atomic distance are skipped. There is no acceleration for either sampling method.