Parametrization of small organic molecules for classical molecular dynamics simulations is
not trivial. The vastness of the chemical space makes approaches using building blocks
challenging. The most common approach is therefore an individual parametrization of each
compound by deriving partial charges from semiempirical or ab initio calculations and
inheriting the bonded and van der Waals (Lennard-Jones) parameters from a (bio)
molecular force field. The quality of the partial charges generated in this fashion depends on …

In the recent publication, 1 we presented a machine learning (ML) approach to predict
partial charges. After publication, we found that the caption of Figure 5 did not contain all
information. For the top panel with external test set 1 (organic liquids), a dielectric permittivity
ϵ= 4 was used, as indicated in the caption of Figure 5 in the original publication, but for the
lower panel with the external test set 2 (FDA approved drugs) ϵ= 78 was used. In order to
confirm the results, we have retrained the ML models with scikit-learn 2 version 1.0 using the …