problem with energyE() calculation?

The Friedel-Crafts reaction in the Reactor Library uses energyE() as the determinant of regioselectivity.

Code:

[bob@epoch bob]$ evaluate -e "energyE" "Clc1ccccc1" 4.05;2.58;2.5;2.54;2.5;2.54;2.5

Note the very small differences between the coefficients of the meta and para C atoms. Yet in nitration reactions of chlorobenzene, the product is about 65% para, 35% ortho, and <1% meta. Friedel-Crafts reactions should give about the same result. So this leads me to ask, is energyE() being calculated properly for haloarenes (using JChem 5.1.0)? If so, is energyE() the best determinant of Friedel-Crafts regioselectivity, or is there a better one?





Note also the fact that energyE() alone gives no ortho/para preference, contrary to observation. To account for this problem, I modify the reaction rules as follows:

Code:

reactivity: energyE(ratom(1)) - min(energyE(reactant(0), filter(reactant(0), "aromaticAtom()"))) <= 0.045 selectivity: -stericHindrance(ratom(1))

The reactivity rule ensures that all atoms within a certain tolerance of the minimum can give products. The selectivity rule orders them by steric hindrance. I arrived at the 0.045 tolerance by experimentation; it works well for everything but haloarenes.





Any insights would be most welcome.

Dear Bob,





Thank you very much for this valuable comment. I will check your rule carefully and test with our molecules.





In general, the EnergyE calculation is simply Huckel based, so might not always be the best measurement for regioselectivity. Luckily, the functions are integrated via a plugin system, so you can incorporate your own calculations instead of EnergyE if you prefer.





You are right, this is just an electronic factor and the effect of steric hindrance should be considered as well. Unfortunately, the sterichyndrance function requires 3D coordinates, and 3D optimization can slow down batch calculations, but I will test the time of your settings too. Alternatively, the stericEffectIndex can be considered too. That is topology based, so very fast, but in my opinion it is not the best for this reaction. I have an alternative topology based method in mind, to specifically handle this case, though its implementation is not yet planned.





On the basis of our recent experiences in biotransformation library design, we plan to incorporate a new feature in reaction rules in this year, that can be used to predict the ratio of regioisomers. It will override the current selectivity rule, and might be used to generate all regioisomers, but with different reaction rates, so the product ratios can be estimated.

Despite the 3D coordinate generation, the reaction is fast with my test set, but not good, because nitrobenzene is acylated with your proposed rules.

Gyuri wrote:

Despite the 3D coordinate generation, the reaction is fast with my test set, but not good, because nitrobenzene is acylated with your proposed rules.

Yes, I add an exclusion rule for Friedel-Crafts acylation.

Code:

energyE(ratom(1)) < 1.1

This rule is not needed for nitration, though the other rules are the same.

Gyuri wrote:

On the basis of our recent experiences in biotransformation library design, we plan to incorporate a new feature in reaction rules in this year, that can be used to predict the ratio of regioisomers. It will override the current selectivity rule, and might be used to generate all regioisomers, but with different reaction rates, so the product ratios can be estimated.

This feature will be very valuable to us.





Even more valuable, though, would be the ability to write stereoselective additions to alkenes (syn or anti), such as OsO₄-catalyzed dihydroxylations, Diels-Alder reactions, etc. We have talked about that in other posts.

Exclusion rule is not perfect, beacuse it throws out the entire a molecule if the rule matches. For example, if a molecule contain an aliphatic amino group, than it will be excluded from the Friedel-Crafts reaction. In your case there can be many ring systems in the molecule, so one can react while the other is not.





The evaluation order:


1. Substructure query with the corresponding reactant of the scheme


2. Exclude rule: if a molecule matches the exclude rule, it is simply skipped. (in our case, such as those having amino, alcoholic hydroxy, etc)


3. Reactivity rule: the reaction sites of the remaining molecules are refined (in our case, checking that sum of partial charges of the matching ring systems)


4. Selectivity rule: if more than one reacitve reaction site is identified, those sites are sorted by the selectivity rule and the site at the top of the lits is transformed according to the scheme. (Tolerance allows to react the top few sites).

Gyuri wrote:

Exclusion rule is not perfect, beacuse it throws out the entire a molecule if the rule matches. For example, if a molecule contain an aliphatic amino group, than it will be excluded from the Friedel-Crafts reaction. In your case there can be many ring systems in the molecule, so one can react while the other is not.

Ahh. In that case, invert it and add it to the selectivity rules.