LSD 3.0

Last modified: July, 8th 2002. Version 3.0.2

Table of contents

From spectra to data

The goal of LSD is to help the user propose one or more molecular structures from data that is mainly extracted from 2D NMR spectra. Recording HMQC (or HSQC), HMBC and COSY spectra is the pre-requisite for LSD use. LSD has no knowledge of chemical shifts. There is no way to introduce this kind of data, for any nucleus type. LSD uses the chemical shift and coupling information provided by the user that are necessary to establish carbon hydridization state or to impose a particular neighborhood to a given atom.

NMR data coding for LSD may initially seem awkward to beginners. The following paragraphs are intended to help them. A useful documentation source may be found in the examples in the Data folder. A data file for LSD is a text file that contains commands, each being followed by one or more parameters. The most useful commands will be first introduced, but a thorough command set description is given in the Data file structure paragraph. A text-type user interface may seem old-fashioned, but it is sometimes easier to modify a text than to navigate though menus and sub-menus whose function is not always totally explicit.

The first step in the elucidation of an unknown molecular structure is to find its elemental formula and to determine the status of all non-hydrogen atoms. The status consists of an index, a chemical symbol, a hybridization state (sp2 or sp3) and a number of attached hydrogen atoms. Defining the status of all atoms is a strong constraint for the user. The status of heteroatoms is not always easy to define, mainly when more than one of their type is present. Even for carbon atoms, the chemical shift value is not sufficient to non-ambiguously deduce the hydridization state. More than a single data set must then be written and tested. More flexibility to status definition will be given in the near future, but for the moment the user has to cope with the present capabilities of LSD.

Atom indexing may be arbitrary. A recommended procedure is the indexing of carbons from 1 in decreasing order of their 13C chemical shift value. Heteroatoms recieve the following numbers as indexes. Carbon indexes are then reported in the 1D projections around the 2D HMQC and HMBC spectra. The command that defines an atom status is MULT. Uppercase letters need to be uppercase. MULT is followed by the atom index, 2 or 3 for the hydridization state (sp2 or sp3) and the multiplicity (number of bound hydrogen atoms). The command MULT 1 C 2 0 says that atom 1 is a quaternary (0) sp2 (2) carbon (C). The data file contains as many MULT lines as non-hydrogen atoms in the molecule.

The following step is indexing the hydrogen atoms. Again, the indexes may be given arbitrarily, but it is handy to give the same index to the proton and to the carbon atom attached to it. The HMQC spectrum matches carbon and proton indexation schemes. Two inequivalent hydrogen atoms within a methylene group get the same index. The hydrogen indexes are then copied in the projections around the 2D COSY and HMBC spectra. A correlation in a 2D spectrum is characterized by the type of spectrum, the index of the atom in dimension 1 (carbon index in HMQC and HMBC spectra), and the index of the atom in dimension 2 (always a hydrogen atom index). The command HMQC 4 4 indicates that carbon 4 and hydrogen 4 are bound together. Commands HMBC and COSY are coded in the same way.

Any easily deduced bond must be given to LSD through a BOND command. It takes the indexes of the bound atoms as parameters. If carbon 1 is that of a keto group and if there is an sp2 oxygen 22, the command BOND 1 22 establishes the bond between atoms 1 and 22.

Unless specified, the order of the commands within a data file has no importance. All command files must have EXIT as the last command. Apart from purely NMR derived commands, other commands are needed for execution control. One should prefer to group them at the beginning of the data file. The most useful one is DISP 1. It makes LSD produce solutions in a format suitable for molecular drawing by the outlsd program. In a data file, one first finds DISP 1, then MULT, HMQC, COSY, HMBC, BOND and EXIT commands. This is the skeleton for basic LSD data files.

A less trivial part in the writing of LSD data files is the definition of atom properties and of the associated lists. An atom property (PROP command) indicates for an atom or for all atoms in a list the exact number of neighbors (1 or more) that must be members of another atom list. The number 0 means "all". A given atom may have only one single property statement. For example, the chemical shift of carbon atoms indexed 8, 9 and 10 is about 15 ppm. The hydrogen atoms indexed 8, 9 and 10 are those of methyl groups appearing at about 1 ppm as singlets. Carbons 8, 9 and 10 are most probably bound only to quaternary carbons. You need to build two lists in order to code this constraint with LSD. The first list contains the atoms 8, 9 and 10 while the second one contains all the quaternary carbons. An atom list is described by its index and its content. The content is either explicitely defined (giving the index of all atoms within the list) or using atom status. It is also possible to combine lists according to the operations defined in the set theory. Back to the example, LIST L1 8 9 10 defines list 1 and puts atoms 8, 9 and 10 in it. QUAT L2 defines list 2 and puts all quaternary carbons in it. Once L1 and L2 are defined, the constraint on atoms 8, 9 and 10 is coded by the PROP L1 0 L2command. It could have also been possible to write PROP L1 1 L2, because a methyl group carbon always has a single neighbor. List and property definition commands are treated in the order they are written in the data file. Changing the order can change the meaning.

A common temptation for beginners is spectra over-interpretation, meaning that non-significant information is introduced. Inconsistent or wrong data leads LSD to fail, producing a message that indicates which is the deepest analysis level reached (the much hated "max stack level: " message). Introducing nJ COSY or HMBC correlations with n greater than 3 also leads LSD to fail. In order to avoid this unpleasant situation, the weakest COSY and HMBC correlations should be first written as comments, by putting a semicolon at the beginning of the corresponding commands. If the number of found solutions is too big, it's time to remove semicolons to impose more property or correlation constraints to the problem.

This brief overview of LSD commands will allow users to be more confident with the writing of data files. However, reading of Data file structure is a mandatory step in the learning of the LSD command language.

From data to structures

A data file named "pinene" is provided in the Data folder.
Running LSD on it is simple achieved using:

myprompt$ cp Data/pinene .

myprompt$ lsd pinene

where myprompt$ is the prompt given by the computer.
LSD writes that 1 solution is found.
A file pinene.sol is created, containing atom connectivity information.
A coordinate file pinene.coo can be then generated:

myprompt$ outlsd 6 < pinene.sol > pinene.coo

The outlsd program creates 2D coordinates from connectivities.
The result is viewed using a Postscript® previewer.
Postscript drawings are first produced with:

myprompt$ genpos < pinene.coo > pinene.ps

and displayed on screen using:

myprompt$ xpsview pinene.ps

if xpsview is the currently available Postscript display program. Remark: The file extensions .sol .coo .mol are not imposed but it is good practice to use always the same conventions.

Very quick start

The command:

myprompt$ solve pinene

does the whole job in one shot: structure resolution, creation and display of the structure drawings.
It is generally not a good idea to proceed this way with a new data set.
However, this can be done once the data set runs from the command line without any error (see From data to structures).

Drawing improvement

The coordinates generated by outlsd sometime produce drawings that are difficult to read.
A simple program named m_edit was written in order to interactively improve the quality of the structure drawings.
You need to have Tcl/Tk and the wish command installed in order to run M_EDIT (version 8.0 or later).
The command

myprompt$ m_edit

launches m_edit.
The File menu allows to read and save files, and to exit from the program. The View menu is used to navigate forward and backward through a set of structures within a file. The supported file formats are .coo (the LSD-specific format) and .mol.

Using m_edit is very intuitive: atoms are moved using the left mouse button.
There is no multiple selection mechanism available yet, sorry.

Checking data

A shell script named chklsd extracts basic molecular information from the data file, such as the number of atoms of each kind, the number of insaturations, double bonds and rings, and the molecular mass:

myprompt$ chklsd < pinene

If generic atoms X, Y or Z are used (see VALE) the molecular mass is correct only if comments
(see Data file structure) like:

;MASS X 19

are provided. In this example X stands for fluorine.

Data file structure

A data file is made of commands and comments.
The number of commands is limited to 300.
A comment is everything between a ";" and the end of the line.

A command is always typed on a single line and is made of a command mnemonic, generally followed by 1 to 4 parameters.
All parts of a command are separated by spaces. Commands are case-sensitive.
The very last command in a data file must be EXIT.

All command mnemonics are made of 4 alphanumeric characters.

Some of them end with space characters (^), such as CH^^.

Mnemonics are followed by parameters, separated by blanks.
The parameter types are described using the following symbols:

The parameters of a command are successively referenced by P1, P2, P3 and P4.
A file must end with the EXIT command (the only one without any parameter).

Basic commands

List definition

The commands that are described here help to define the lists of atoms required by the PROP command.
They are interpreted in the order that they are written.
They are sorted according to their number of arguments :

Execution control

Substructural information

Solutions found by LSD after correlation analysis and uncomplete atom pairing can be eliminated if they do not contain a given substructure.
Only one substructure is allowed. It may be made of non-connected pieces.
A substructure is a set of sub-atoms connected by sub-bonds.

Result processing with outlsd

In order to use outlsd, LSD must be run with DISP 1.
The outlsd program reads from and writes to standard devices.
It takes an integer n comprised between 1 and 7 as argument.

myprompt$ outlsd 7 < pinene.sol > pinene.mol

Depending on n, the result contains

Option 2 requires some unavailable proprietary software.

Result viewing with genpos

The option 6 of outlsd produces an output that starts with the magic word DRAW, in a format suitable for its treatment by the genpos program. genpos reads and writes from standard devices and generates PostScript instructions.

myprompt$ genpos < pinene.coo > pinene.ps

The resulting file can be opened by Postscript previewers or be sent to Postscript printers.