LSD as seen by ACD

What is LSD (in this context)?

LSD (Logic for Structure Determination) is a piece of computer software that was written in order to help organic chemists to propose molecular structures of unknown substances from 2D NMR data sets.

Everything about LSD here.

What is ACD?

ACD (Advanced Chemistry Development) is an international private company that delivers chemistry computer software tools. ACD sells a structure determination software named "StrucEluc".

Why was this web page written?

This web page was written by the author of the LSD software after the publication of the paper
entitled "Computer-assisted structure verification and elucidation tools in NMR-based structure elucidation"
written by M.E. Elyashberg, A.J. Williams, and G.E. Martin
and published in "Progress in Nuclear Magnetic Resonance Spectroscopy" Vol. 53 (2008) pp. 1–104.

The first two authors are ACD employees, while the third one works for the Schering-Plough Corporation, Pharmaceutical Sciences.

In this review article, the authors report what is written in the scientific literature about the LSD software, as one can expect from a review article. Some recent developments of LSD seem not to have been taken into account. Minor technical or general points also need to be clarified. These are the reasons for which this web page was written.

In the following section, citations from the orginal article appear in red .

Off we go...

The encoding of data for use by the LSD program is a manual process requiring careful inspection of the plotted spectra.

Yes, data encoding is manual. At least, this features makes it NMR-software-independent. Chemist still have the habit (bad or good?) to draw spectra on large-size sheets of paper in order to have a clear and general overview of their data. A computer screen is sometimes just too small to be of practical use. However, recent efforts towards computer-aided data encoding for LSD should be made visible shortly.

Yes, you have to be careful. Who would say: "you haven't to be careful"?

all suspect data is discarded to avoid the risk of introducing uncertainties into the final analysis.

I hope that everyone, independently of which software is used, will follow this advice. Chemists know that it is not always easy to take a decision about what should be done about a particular HMBC correlation spot.

Nuzillard et al. have not revealed the criteria that are applied to define the status of the heteroatoms. In those situations where the hybridization state is not easily determined, for example where carbon nuclei resonating around 100 ppm can be either sp2 or sp3 in nature, then separate data sets must be constructed to take into account the various possibilities that can exist with regard to the atom status.

There is simply no criterion, and therefore nothing to reveal. Why should I make secrets? As correctly written by the authors, the user has to try the various status possibilities for the heteroatoms. This is a limitation of LSD that has to be broken.

Properties deduced from the analysis of spectral data i.e., chemical shifts, spin–spin couplings etc., are also used as inputs to specify the status of neighboring atoms and thereby restrict the sets of possible atom-atom pairings that will result from the bond generation process.

Not exactly. Chemical shifts and spin-spin couplings are simply used to restrict the list of the possible neighbors of an atom. The lists may be defined using parts of the status, but by other means as well, including simple enumeration.

Only one substructure is allowed. Nuzillard et al. did not explain how they define the chemical shifts of the sub-atoms included into the structural constraints. 13C signals may be assigned on the basis of spectral data referring to previously studied structures.

The first sentence is now incorrect. You may define as many sub-structure as you want. Their definition have to be stored in seperate files. A LSD command allows the user to specify how sub-structures can be combined together in order to obtain the desired solution filter. The following cited sentences are more obscure. The question could be: "How are the sub-structure(s) chosen by the user?". The answer is: "By any legal mean.".

In our opinion the input of NMR data to the LSD program is awkward since a data entry file is a text file containing commands with each followed by one or more parameters.

The typing of a human readible text is not that awkward, contrarily to what authors would let to believe. Again, the LSD user is not bound to a particular data processing software. The 4-letter command names are not so difficult to remember. Modern text typing software also have cut'n paste shortcut keys.

In addition, defining the status of all atoms is a strong constraint for the user

Yes, that's true, but readers are not all supposed to suffer from Alzheimer's disease.

Any easily deduced bond must be introduced to LSD through a special command.

"Must" is wrong. You do it if you want to do it. I you are not sure, you don't do it. This philosophy is general in LSD and most probably in other software. The "special" command name is "BOND". This is rather handy to indicate the existence of a bond. Why should this command be "special"?

An invalid correlation can be eliminated to check if it is responsible for the program failure; consequently, the detection of invalid correlations is a problem shouldered by the user, which is not the most efficient approach by any means.

This was true in early versions, but has become wrong now, as explained by the authors themselves, just immediately after this sentence. I do not see the point in writing wrong things when you write yourself that they are wrong.

A special command provides the maximum number of correlations that it is possible to eliminate so that at least one solution is produced.

A "special" command, again. Well, a command... that partly solves the problem of very long-range HMBC correlations, contrarily to what was written immediately before.

The substructure search mechanism in the LSD program may not be powerful enough to eliminate cyclopropenes or cyclobutadienes, but a filter is able to retain from the result file structures that do not contain 3- or 4-membered rings and they are finally stored.

The solution filter in LSD is really able to eliminate cyclopropenes or cyclobutadienes, if you want to do it. Users are allowed to write their own sub-structures definition files for solution filtering.

The examples reviewed here demonstrate that while the data input process has the potential to be somewhat daunting, in the LSD system (using command settings, setting skeletal atom hybridization, and the possibility of the presence of neighboring heteroatoms is obligatory), the system is nevertheless still quite capable of analyzing and solving complex chemical structures.

Daunting? Hopefully, organic chemists are not chicken-hearted people. What could be daunting in the field of automatic structure elucidation would be the price of StrucEluc and of the required software suite. After the repetition of the weak points of LSD, the authors finally admit that LSD could eventually be used to solve non-trivial problems. Fine!

The system is available for testing via the internet.

Not only for testing! You may use it. LSD is free software, "free" like in "free speech", and in "free beer" as well. I read once on the web (the page is no more available) that someone used LSD "as a free alternative to StrucEluc".

If you want to use the LSD software and if you have questions, I will try to answer them.
Dr. Jean-Marc Nuzillard.

May 4, 2008.