Journal: Journal of Computational Chemistry

Article Title: Evaluating the Functional Importance of Conformer‐Dependent Atomic Partial Charge Assignment

doi: 10.1002/jcc.70112

Figure Lengend Snippet: Differences between partial charge generation methods. Different charge assignment protocols/implementations result in different levels of conformer dependence for assigned partial charges, depending on the input conformer(s) used for partial charge generation. We generated 50 partial charge sets of each molecule using 4 different partial charge methods: Antechamber AM1‐BCC, OpenEye AM1‐BCC, OpenEye AM1‐BCC ELF10, and OpenFF NAGL. (a) Using maximal partial charge difference as our metric, Antechamber AM1‐BCC charges on average showed less variation than the OpenEye equivalent AM1‐BCC charges, and OpenEye AM1‐BCC ELF10 charges showed slightly less variation on average. As expected, OpenFF NAGL showed no conformer dependence. (b) Using our second metric, maximal Δ q _ b o n d difference, OpenFF NAGL again showed no conformer dependence. Again, on average, OpenEye AM1‐BCC ELF10 charges showed less variation on average as compared to Antechamber and OpenEye AM1‐BCC charges.

Article Snippet: Conformers were generated starting from 50 different random seeds to generate 50 sets of 500 random conformers, the minimum required for OpenEye AM1‐BCC ELF10 charges [ ].

Techniques: Generated

Journal: Journal of Computational Chemistry

Article Title: Evaluating the Functional Importance of Conformer‐Dependent Atomic Partial Charge Assignment

doi: 10.1002/jcc.70112

Figure Lengend Snippet: In 2 out of 3 pictured molecules, l i g _ 215 and l i g _ 25 , we see smaller ranges of Δ q _ b o n d differences when charging with OpenEye AM1‐BCC ELF10 as opposed to AmberTools AM1‐BCC. In these molecules, we see a smaller range of Δ G using the less variable charging method, OpenEye AM1‐BCC ELF10. In l i g _ E 27 , OpenEye AM1‐BCC ELF10 and AmberTools AM1‐BCC have a similar maximal range of Δ q _ b o n d difference, but OpenEye AM1‐BCC ELF10 charging results in fewer bonds with variability above 0.02 e.

Article Snippet: Conformers were generated starting from 50 different random seeds to generate 50 sets of 500 random conformers, the minimum required for OpenEye AM1‐BCC ELF10 charges [ ].

Techniques:

Journal: Journal of Computational Chemistry

Article Title: Evaluating the Functional Importance of Conformer‐Dependent Atomic Partial Charge Assignment

doi: 10.1002/jcc.70112

Figure Lengend Snippet: Comparison of calculated AHFE Δ G values using OpenFF NAGL and OpenEye AM1‐BCC ELF10. NAGL and AM1‐BCC ELF10 charges produce similar average Δ G values across repeats; 5 out of 6 molecules are under 1.5 kcal/mol different. Using consistent partial charges, such as through a conformer‐independent method like NAGL, results in less variability across repeats as compared to AM1‐BCC charges alone. However, the variability with NAGL is similar to that observed with the less conformer‐dependent OpenEye AM1‐BCC charges. Since NAGL doesn't have variable charges at all, this indicates a background level of statistical error that is present in all calculations.

Article Snippet: Conformers were generated starting from 50 different random seeds to generate 50 sets of 500 random conformers, the minimum required for OpenEye AM1‐BCC ELF10 charges [ ].

Techniques: Comparison