Platt and Randic with or without explicit H?

Hi,


is the Platt index defined with/without adding explicit Hs?


This paper lists 12 for the Platt index [PDF].





Marvin either calculates 21 or 96


Platt index 12 96 for C8 (Octane, CCCCCCCC)


----





Another one is the Randic descriptor


Randic index 3.91 10.75 for C8 (Octane, CCCCCCCC)


either without and with H.





Dragon always adds explicit Hs.





Tobias

Hi,





so does our TopologyAnalyser: both in case of Platt index and in case of Randic index explicit hydrogen atoms are considered.





Regards


Miklos

Hi Miklos,


assume you have a file with hundreds of mixed molecules


some with H and some without H. My simple


example was only for one compound.





It could be in principle coded in SMILES and SDF.





Marvin TopologyAnalyzer - Octane without H


CCCCCCCC


Platt index = 12


Randic index = 3.91





Marvin TopologyAnalyzer - Octane with H


[H]C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])[H]


Platt index = 96


Randic index = 10.75





So Marvin and potentially cxcalc calculate wrong Platt and Randic indices.


Otherwise the TopologyAnalyzer has to turn


on explicit H by itself. But I am not sure if by definition


those indices require H or not.





Cheers and thanks


Tobias

Hi Tobias,





let's clarify the situation:





1. You refer to a paper that states that the Platt index of octane is 12. Then you say that marvin calculates either 21 or 96. However, in your second post you say it's 12 that marvin calculates.





2. Regarding hydrogen atoms: I think I misunderstood you when you said that 'Dragon always adds explicit Hs'. I though you meant that it considers them in the calculation of the platt index. And when I said 'so does marvin' I meant that, namely: marvin also considers explicit H atoms (but only those, not implicit ones!!!) when calculates the platt index.





But now, reading your second post I understood that you meant that Dragons 'hydrogenizes' the structures before calculating the platt index.


Our calculators do not do that for many reasons. If you think that taking all Hs into account in calculating the platt (or other) index is needed then you need to apply the required standardization rules (or simple hydrogenize) the structure. This is important if you consider the importance of pH, protonation, tautomerization etc. The calculator is not aware of these circumstances or the users intention, it performs calculation on the exact structure passed to it without doing any transformation on the structure.





Does this explanation help?





Miklos

Hi Miklos,


for logP (physicochemical descriptor) I get the same results


adding explicit Hydrogens or not. Seems Correct. Most of the available 2000 descriptors will be useless anyway and will be removed by feature selection, but the calculations should be


consistent.





For Platt, John R.; Influence of Neighbor Bonds on Additive Bond Properties in Paraffins; J. Chem. Phys. 15, 419 (1947); DOI 10.1063/1.1746554 and a nice description also here Prediction of Isomeric Differences in Paraffin Properties; John R. Platt; J. Phys. Chem.; 1952; 56(3) pp 328 - 336; DOI: 10.1021/j150495a009 and just for reference Topological indices for structure-activity correlations DOI 10.1007/BFb0111207 they were not very helpful, but looking in some books on Google Books it seems that those indices including Platt and Wiener work on the hydrogen surpressed graph





Quantitative Structure-activity Relationship (QSAR) Models of Mutagens and Carcinogens (Romualdo Benigni) also writes: The adjacency and distance matrices correspond to the hydrogen-suppressed graph. So I would just follow the mainstream and calculate the Platt and Wiener index without explicit H and not letting it to the user in this case, who may not even be aware of. [LINK].





So I will just update my code using the CalcDescSimple (Chemaxon Forum) but


for the MARVIN plugin it would be good too to use the hydrogen surpressed graph for the Platt and Randic index.





In case of cxcalc and Marvin the user woul dbe on the safe side, in case of the API he may or may not calculate it in the wrong way using Hs.





Cheers


Tobias

Thank you, Tobias, for pointing that problem out, TopologyAnalyser will always return the results for hydrogen surpressed graphs. I am examining the differences of the values now.

Hello,


thank you. Are there any plans for extension of number of possible descriptors or is this something for 3rd party implementations or low priority.


Tobias

There are some descriptors that could be added, though we are not sure about their importance in practice. If you think, that an important one is missing, just please send me a proposal for implementation.





Anyway, yes, we are planning to extend the current descriptor set of the TopologyAnalyser plugin, for example, new shape descriptors will be added in the near future.

TopologyAnalyser will continue calculating topological indexes using the nodes of the molecule graphs. So explicit hydrogens will be considered, implicit hydrogens will be ignored. In Marvin, you can convert hydrogens easily before the calculation.





You can access the same hydrogen handling functions in the programming API or you can batch convert files using MolClonverter or Standardizer before calculations. We will additionally introduce a new Chemical Terms function "standardize" that wil allow you to convert the hydrogens of the molecules even when you calculate the indexes from Chemical Terms.