Thresholds

CENSO is a threshold based sorting algorithm. The thresholds have been determined for typical drug-like organic molecules up to 200 atoms. For cases with high flexibility or very different conformer ordering between SQM/FF and DFT it can be necessary to increase the thresholds. If your target quantity is an averaged ensemble free energy, be sure not to use small energy windows in part0 and part1, as this can falsely reduce the ensemble size.

In the following the employed thresholds are listed and their use is explained:

Part0 - Cheap Prescreening

part0_threshold = g_thr(0)
cml:

-part0_threshold

censorc:

part0_threshold

part0_threshold g_thr(0) is an energy window /threshold in kcal/mol within which all conformers are considered. The CENSO internal default for g_thr(0) = 4.0 kcal/mol. The sorting threshold is designed to remove conformers very high lying in electronic energy (E).

Part1 - Prescreening

In part1 two sorting steps are applied. One based on g_thr(1) with improved energy and solvation description and one on G_thr(1) where thermostatistical contributions have been calculated additionally.

part1_threshold = g_thr(1) and G_thr(1)
cml:

-part1_threshold

censorc:

part1_threshold

part1_threshold g_thr(1) is an energy window /threshold in kcal/mol within which all conformers are considered. The CENSO internal default for g_thr(1) = 3.5 kcal/mol. The sorting threshold g_thr(1) is designed to remove high lying conformers based on improved electronic energy (E) and solvation contributions (δG_solv). G_thr(1) sorting is based on full free energy including thermostatistical contributions (G_mRRHO). Both, g_thr(1) and G_thr(1) use the same base threshold e.g. 3.5 kcal/mol and G_thr(1) can be automatically increased as a function of the standard deviation of G_mRRHO in the structure ensemble, indicating high structural diversity and possibly larger errors (fuzzy sorting).

Part2 - Optimization

In part2 two sorting thresholds are applied. One which is applied during geometry optimization and one Boltzmann sum threshold.

threshold applied during optimization G_thr(opt,2)
cml:

-opt_limit

censorc:

opt_limit

Spearman-threshold for testing for parallel PES
cml:

-spearmanthr

censorc:

spearmanthr

Boltzmann sum threshold G_thr(2)
cml:

-thrpart2

censorc:

part2_threshold

The internal default for G_thr(opt,2) is set to 2.5 kcal/mol. During the geometry optimization the initial threshold G_thr(opt,2) is increased by 60 % (or at least 1.5 kcal/mol). If the PES can be assumed parallel (tested by a Spearman rank coefficient) the threshold is decresed until it reaches the initial value. All conformers with (free) energies above the G_thr(opt,2) threshold are discarded and all conformers below G_thr(opt,2) will by fully DFT optimized. The last threshold employed in part2 is a Boltzmann sum threshold G_thr(2) and all populated conformers of the ensemble, up to the Boltzmann population (in %) are considered further.

Part3 - Refinement

In part3 a Boltzmann sum threshold is employed.

Boltzmann sum threshold G_thr(3):
cml:

-thrpart3

censorc:

part3_threshold

Based on high level free energies Boltzmann weights are calculated and all conformers up to the Boltzmann sum threshold (in %) are considered further.

Part4 - NMR properties

In part4 only Boltzmann sum thresholds of part2 G_thr(2) or part3 G_thr(3) are applied. It is possible to calculate part4 using Boltzmann weights from part1, but since there is no Boltzmann sum threshold in part1, in this case, the threshold G_thr(2) is used. The NMR properties are calculated on populated conformers up to the Boltzmann threshold in %.

Part5 - Optical rotatory dispersion

In part4 only Boltzmann sum thresholds of part2 G_thr(2) or part3 G_thr(3) are applied. It is possible to calculate part5 using Boltzmann weights from part1, but since there is no Boltzmann sum threshold in part1, in this case, the threshold G_thr(2) is used. The OR property is calculated on populated conformers up to the Boltzmann threshold in %. For optical rotation it is necessary to include almost the entire ensemble e.g. 99 %