The system training algorithm owes its power and accuracy to the automatic elimination of all the fundamental problems related to any fragmental method of calculation:
- Missing fragmental increments in the algorithm. ACD/LogP uses
your experimental data in order to calculate the missing parameters.
- Incorrect fragmental increments. The algorithm may contain
incremental values derived from unreliable experimental data. ACD/LogP replaces them with your data.
- Missing correction factors. For example, intramolecular H-bonding
is usually taken into account only in rigid cyclic systems, but rarely in non-rigid acyclic skeletons.
ACD/LogP automatically splits your structures into such fragments that are not affected by the missing correction factors.
- Error accumulation in complex molecules. Summation of dozens of
fragmental increments and correction factors may lead to a large error even if all parameters are fairly accurate.
ACD/LogP always keeps down the number of required parameters by using the most similar structures from your
database.
- Incorrectly drawn structures. Many important structures exist in
multiple ionic or tautomeric forms and it is not always clear how to draw the structure in order to obtain the
correct result. ACD/LogP overcomes this difficulty by finding the most similar structures from your database.
So, if you consistently draw your structures in the same fashion you will always obtain correct calculations.
|
