How to Search for Conformers in Spartan
Help File from Wavefunction
Frequently Asked Questions
* How do I know I have the global minima?
* Is the systematic algorithm reproducible?
* Is the global minima the truth?
* How should I use QM methods with Conformational Analysis?
* What are the details of the algorithm.
Systematic description
Monte-Carlo description
Moves
Minimization algorithm
* How can I modify the algorithm?
* What do these columns mean in the output file?
* Why does the output say it is removing molecules from the list and how is it deciding what to remove?
How do I know I have the global minima?
The only way to be sure you have the global minima is to do a systematic search with a very high grid. For example, change the default for a SP3-SP3 carbon bond rotor from the default of 3 to 10 (or more if the energy surface is unusually complex). This is known to be very inefficient and chances are that if you have a non-trivial molecule you may not have the patient to do this.
Our experience with medium size organics is that a simple monte-carlo algorithm usually does a good job of finding the global minima. See "How do I know my MC result is good?", and "Is the global minima the best minima?"
Is the systematic method reproducible?
Yes, If you start from the same original conformation. In simple molecules systematic is almost always inclusive of all the (important) minima. As molecules get more complicated the default grid size, while likely correct in finding all classes of minima may skip some minima. For example the single [-120,+120] conformation of olefins might bifricuate into multiple minima as steric bulk increases. It is likely that one of the two new minima would be found using the standard systematic approach. But which one would depend sensitively on the initial conformation.
To insure you find all these one might want to increase the default rotor value. Of course doing this has a great cost in computational time, and a better approach would be to switch over to the monte-carlo method.
Another example of the initial starting conformer affect the result can be found by looking at a 8 member carbon chain; "C1-C2-C3-C4-C5-C6-C7-C8". For illustrative purposes let's just look at the central "C4-C5" bond as we rotate it 360 degrees; from -180 to 180 degrees. If we start in the all trans conformation, we find that as we rotate the "C4-C5" bond the final conformation of 360 degrees is identical to the initial at 0 degrees. However, if the initial conformation had a number of kinks in it, we might discover that at the 120 degree mark, the C1 and C8 ran into each other. To relieve this steric problem the other dihedral angles, would relax, likely change by more than 100 degrees, falling into a new well. As we continue the coordinate driving of the central C4-C5 angle to trans (180), we might find that the final conformation is not the same as the initial conformation because these other dihedrals have changed.
Is the global minima the truth?
It is often the most interesting. And at the least, is often representative of a conformer found at room temperature. However, what conformation is the best depends on what you are looking at. There are often many other variables which are ignored in the conformational analysis, including effect of solvent, level of theory.p
How should I use QM methods with Conformation Analysis?
Because QM methods take so much more time than mechanics, (by orders of magnitude) it is usually a mistake to try the conformational algorithms with these methods. One usually uses the MMFF mechanics force field to generate a list of low energy conformers. This list is then resubmitted at the desired level of theory as either a "geometry optimization" or a "single point". The original MMFF conformers usually changed geometry slightly and the relative energies may differ.
For very small systems (1-2 degrees of freedom) it is sometimes possible to use conformational analysis to scan conformer space. However, the time required, as well as the possibility of bad initial conformers with steric problems causing the breaking and forming of bonds require the user to be careful when using QM and conformational analysis. ?
What are the details of the algorithm?
The conformation module has two modes:
* systematic (SYSTEMATIC).
* monte-carlo (MONTE-CARLO) and
Each of these algorithms consist of moving or rotating one or more of the molecule's bonds, followed by a minimization. Each of these topics are discussed in turn, followed by a short discussion of how to customize the algorithm and keywords
SYSTEMATIC uses a systematic method to explore the vast majority of conformations of a molecule. For acyclics, the method is simply to rotate each bond by a specified angle (usually 120 degrees) and search for conformations. This typically spans the space effectively enough to find all conformations of small molecules. For cyclics, the same method is used to sequentially bend rings within the molecule. However, the conformations of rings are not necessarily well defined, and this may not result in a perfect search. For large molecules, this method is very time consuming, since the number of conformations to search grows exponentially. For this reason, SYSTEMATIC should only be used for small, preferably acyclic, molecules.
MONTE-CARLO uses a "simulated annealing" method to generate conformations of a molecule. The idea here is to randomly rotate bonds and bend rings within the molecule until a preferential (minimum energy) geometry is attained. Initially, the molecule is considered to be in a high-temperature system; this means it that has enough energy and is thus flexible enough to move from a low to high energy conformations. This is important because often the global minimum configuration of a molecule may be very different from the initial conformation. As more conformations are explored, the temperature of the system decreases, making the molecule less inclined to move out of low energy conformations, and thus looking more closely at other minima in the currently nearby vicinity.
The algorithm is a standard simulated annealing algorithm with a temperature ramp of
T = T(final) + K*(1-I/Imax)^3
Some modifications have been done to avoid dead-ends; if the system appears stuck, the current conformation will be replaced with conformer a randomly picked, but previously calculated, conformation. The new conformation is weighted via the usual (normalized) Boltzmann criteria.
Moves:
The basic moves are the same for both systematic and monte-carlo algorithms:
1. Basic torsion rotation:
The basic move in conformational searching. This can have some undesired long range effects, such as greatly straining (or breaking) rings.
2. Osowa wag:
2 correlated rotations which keep ring closure. If 4 atoms (A,B,C,D) are connected in series, the atom C, and everything connected to it is connected to rotated around the B--D axis. (Some rearrangement of the spinnage on atoms B and D are required.) This move is the default move for any atoms in a ring.
3. 6-member flip: For 6 member rings, if 2 (and only 2) opposite atoms are selected they are flipped in pairs. This move switches from one chair conformation to another, thus making it rather unlikely that one will find a twist-boat conformation (which is usually much higher in energy). If one wants to catch these conformations she should select more atoms in the six member rings (and increase the WINDOW variable).
Minimization: For mechanics runs a multi-step minimization algorithm is used:
1. A rigid rotation (or set of rotations) is applied the current molecule. (see moves)
2. A minimization is applied given the set torsion constrains.
3. The constraints are relaxed and
4. a second minimization is applied.
5. The constraints are removed,
6. A fast minimization is done to get the rough geometry
7. If the energy is too high, or the geometry has diverged from the goal geometry by more than 90 degrees the job is considered to have failed.
8. A final minimization is done to remove all residual forces on the molecule.
If a semi-empirical or quantum mechanical conformation is requested, steps 1 and 2 listed above are completed, followed by a minimization at the requested quantum mechanics level. As mentioned in
How to customize the algorithm
Within the conformer panel, you can choose the bonds and ring atoms to be involved in the conformational search. This can be done by double-clicking on either the bond or atom within a ring. When a bond or atom is selected for rotation, a type-in box will appear in the conformer panel. This box is used to enter the number of increments in the rotation. For instance, if the value of 3 is entered, rotations of 0, 120 and 240 degrees will be applied. If you do not select any bonds or ring atoms the module will use heuristics to decide which elements are relevant in attaining new conformers. Unless you have some chemical insight as to the relevant rotatable members, leaving the selection up to the module is a good idea.
Keywords:
Keywords specific to the conformational module
SEARCHMETHOD=
SYSTEMATIC to force the systematic algorithm
MONTECARLO or MC to force the Monte-Carlo algorithm.
SYSTEMATIC for small systems
MONTECARLO for big systems
MAXCONFS=
...
100
WINDOW=
...
10.0 (kcal/mol)
MAXITR=
...
complicated function
STARTTEMPERATURE=
The initial temperature for the monte-carlo/simulated-annealing algorithm.
5000
PRINTLEV=
Control Printing. See What does these columns mean in the output file?
PRINTLEV=2 displays a label for each conformation.
PRINTLEV=3 dumps the intermediate minimization output to the main output window
Description of the output file
Description of the columns:
* column 1: Iteration
* column 2: Energy of current conformer
* column 3: Current MC temperature
* column 4: Local success rate. Averaged over the last few iterations.
if (*) appears the current conformation was accepted by the MC search algorithm as the basis for the next conformation.
* column 5: Remark, Some extra comments about the conformation:
o "New Best: The current lowest energy
o "Duplicate: Minima has been previously found
o "Chirality: Minimization failed, chirality has changed
o "No Minima: Torsions have deviated too much from goal
o "Strained": Highly strained (unphysical) conformation
* column 6: Conformation label. Only available with PRINTLEV=2 or greater.
Labels each conformation by breaking each torsion into 12 classes each spanning 30 degrees:
*
* c Cis c : -15 .. 0 .. 15
* 0.0 c+ c+ : 15 .. 30 .. 45 c- ~ -30
* -30 | 30 g+ g+ : 45 .. 60 .. 75 g- ~ -60
* -60 \|/ 60 p+ : 75 .. 90 .. 105 p- ~ -90
* -90 -----*----- 90 p+ l+ : 105 .. 120 .. 135 l- ~ -120
* -120 /|\ 120 t+ : 135 .. 150 .. 165 t- ~ -150
* -150 | 150 l+ Trans t : 165 .. 180 ..-165
* 180 t+
* t
Why does the output say it is removing molecules from the list and how is it deciding what to remove?
Any conformer which has an energy greater than WINDOW (10.0) of the lowest energy conformer is thrown away. If there are more than MAXCONFS (100) conformers with acceptable energy will the program will begin throwing away conformers with the goal of keeping as disperse of a group at the same time as keeping the lowest energy conformers.
information provided by,
Sean Ohlinger
Wavefunction, Inc.
Customer Support
.....................................
18401 Von Karman
Irvine, CA 92612 USA
949-955-2120 phone
949-955-2118 fax
www.wavefun.com
confsearch_spartan.doc, 10 Jan. 2003