1147-2-uo2a4-LigandScoringProtocol-Back-Propagation-CatS.txt

Name

CNDO-EEVA-BackPropagation

Software

OpenBabel 2.3.90
CNDO Program, Pople and Bevridge (1970) Fortran Code in "Approximate Molecular Orbital Theory", McGraw-Hill, New York, 1970.
GNU Fortran (Ubuntu/Linaro 4.8.2-19ubuntu1) 4.8.2
In-house PLS regression code

Parameters

N/A

Method

A QSAR model, based on the GC3 CatS dataset as a training set, was developed using
the EEVA descriptor approach of Tuppurainen and Ruuskanen. The 3D geometries of the ligands were created
with OpenBabel 2.3.90 from the provided SMILES strings, and optimized with the MMFF94 forcefield (n=20000 steps)
with the Oboptimize program. A modified version of the original Pope and Bevride Fortran program was coded to create EEVA descriptors from molecular orbital energies.
We used this CNDO-EEVA code to create EEVA descriptors for each ligand in the GC3 and GC4 sets, using CNDO molecular orbital eneriges calculated from the MMFF94-optimized ligand geometries.
The molecule EEVA descriptors from the GC3 CatS set were combined with corresponding IC50 values to make a training set for a simple back-propagation single-layer neural network. The network was trained using 11 hidden neurons and
one output neuron, with an initial learing rate of 0.1 for all neurons. This neural network was used for predicting
IC50's for the GC4 target set.
EEVA REFERENCE:
Kari Tuppurainen, Juhani Ruuskanen,
Electronic eigenvalue (EEVA): a new QSAR/QSPR descriptor for electronic substituent effects based on molecular orbital energies. A QSAR approach to the Ah receptor binding affinity of polychlorinated biphenyls (PCBs), dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs),
Chemosphere,
Volume 41, Issue 6,
2000,
Pages 843-848,
ISSN 0045-6535,
https://doi.org/10.1016/S0045-6535(99)00525-1.
=======

Answer 1

No

Answer 2

Yes