cutoff  type={MIN | MAX}  \
    level=<cutoff level>  \
    [smooth=<NONE | QUADRATIC; defaults to NONE>]  \
    [field_list=<comma/hyphen separated list | ALL; defaults to ALL>]


The cutoff keyword is used to set a threshold value which variables cannot exceed. Variables belonging to fields specified in field_list which have values above (if type=MAX) or below (if type=MIN) the threshold are forced to assume the threshold value. Its purpose is avoiding that a few high values in the dataset may severely bias the model. For example, the GRID program by default sets a threshold value of 5.0 kcal/mol for field energies, while in CoMFA by default energies are limited to 30.0 kcal/mol. If a QM software is used to compute, e.g., electrostatic potential (ESP) fields, a threshold should be set as well, or grid nodes near the nuclei where the ESP assumes high positive values will dominate the PLS model. If a somewhat smoother cutoff paradigm is desired, the smooth parameter may be set to QUADRATIC; in this case energy values will be treated as follows [1]:

let J = 1.2 * level
let K = 0.8 * level

Energy values greater than J in absolute value will be set to level just as with smooth=NONE. On the contrary, energy values whose absolute value lies in a range between 0.8 * level and 1.2 * level assume the value determined by the following quadratic function:

Xcutoff = 1 / (2 * (K - J)) * (X2 - 2 * J * X + K2)

where X are the original energy values and Xcutoff are the energy values after the cutoff procedure has taken place.


# the following command sets a MAX cutoff of 10.0 kcal/mol on grid values belonging to field 1, so that all values exceeding 10.0 kcal/mol will be set to 10.0 kcal/mol
cutoff  field_list=1  type=MAX  level=10.0

# the following command sets a MIN cutoff of -30.0 kcal/mol and a MAX cutoff of 30.0 kcal/mol on grid values belonging to field 2, using a quadratic smoothing of the cutoff
cutoff  field_list=2  type=MIN  level=-30.0  smooth=QUADRATIC
cutoff  field_list=2  type=MAX  level=30.0  smooth=QUADRATIC


  1. Cramer, R. D., III; DePriest, S. A.; Patterson, D. E.; Hecht, P. In 3D QSAR in drug design: theory, methods and applications; Kubinyi, H., Ed. 1993, ESCOM Science Pub., Leiden, pp 456-457

Print version
Mailing list

Last update:
April 18. 2018 12:00:30

Powered by
CMSimple - CMSimple-Styles

Get Open3DGRID at Fast, secure and Free Open Source software downloads

Would you like to align your
dataset? Try Open3DALIGN
Wish to compute a 3D-QSAR
model? Try Open3DQSAR