
22 Patenting Evolved Bactericidal Peptides
526
~
,
Forward modeling: This refers to the use of linear or nonlinear modeling
methods such as partial least squares (PLS) or neural networks (NN) to
predict molecular properties, often called quantitative structure activity re-
lationships (QSARs). These methods build predictive models based on ex-
perimental data. They may use molecular parameters derived from the 3D
structure of the peptide, or a structural description of the molecule (Venka-
tasubramanian, Chan, and Caruthers 1994, Venkatasubramanian et al. 1994).
Model inversion/optimization