LILLY - Stacks

LILLY RESEARCH LABORATORIES

DIVISION OF

ELI LILLY AND

COMPANY

. INDIANAPOLIS, INDIANA 46206 "

TELEPHONE

(317) 636-2211

January 11, 1979

Dr. Dennis H. SmithDepartment of ChemistryStanford UniversityStanford, California 94305

Dear Dennis:

As Ray Carhart can tell you, I am always pleased to offersuggestions and criticisms of the CONGEN program. The onlyproblem is that I will probably give you much more feedbackthan you really need.

I believe that the CONGEN program is a very significant con-tribution to the field of structure elucidation. I have usedit on numerous occasions in the past, and I have grand plansfor the future. Aside from its other more obvious virtues,the program has the potential of being used as a "laboratorynotebook." In my experience, structure elucidations do notfollow a simple, straightforward course. Frequently, theproblem can lie dormant for months, only to be resurrectedwith a burst of new activity. I am hopeful that the CONGENfiles created at each step along the way will provide usefulreminders of the current status of a problem.

Regarding the main portion of the program, I have few comments.I had been generally familiar with CONGEN and the way it didthings, and aside from learning a new command structure, Ihad very little trouble getting started. I was particularlypleased at the speed of the new version. While such com-parisons are subject to differing load averages and faultymemory, I would estimate that the new version is up to tentimes faster than the old, at least at certain operations.

After over two years of using CONGEN on "real life" problems,I can state with some confidence that the overwhelming problemwith the program is that it thinks of far too many ways toput molecules together. It is not uncommon to find oneselfconfronted with thousands of candidate structures, and itseems to me that the best possible opportunities for improvingthe program lie in efforts designed to alleviate this problem.In interpreting any set of CONGEN results, the first difficultyis simply that of visualizing the structures. I have spent

;

January 11, 1979Dr. Dennis H. Smith 2

many hours "translating" CONGEN drawings to something moreuseful to human beings. It is obvious that any improvementsin the drawing program would be extraordinarily useful.However, I believe that it is necessary to recognize that,even as it stands, the structure drawer is a remarkableachievement. A great deal of work might be expended tomake only small, cosmetic improvements. I think it is

relevant, however, to note an approach which might beeasily implemented. At the workshop, Ray mentioned thatthe structure drawer used a pseudo-three-dimensionalrepresentation of the molecule. Could one simply rotatethis representation around the three principal axes beforethe drawing is converted to two dimensions? One might _ thereforeadd a command UNWIND, which would carry out such rotations, andthen redraw the structure. The user could request, for example,that structure #234 be "unwound" in an effort to get something

he could better visualize. If the new drawing was no improve-ment, he could try again until a reasonable drawing resulted.

SURVEY is another command with great potential for dealing withlarge numbers of candidate structures. I have not used SURVEYmuch, so perhaps the criticisms below are unwarranted. But Idon't understand why you decided to use a library file as thesource of pattern structures. If I am examining a file of 1,374structures, I might want to know how many of them have cyclo-propane rings. The obvious thing to do is get into EDITSTRUCT,

build a cyclopropane, and then run a SURVEY. As it is now, I

have to SAVE, EXIT, RUN CONGEN and create a new library file,EXIT, RESTORE, and then SURVEY. It seems to me that one wouldprefer to use the same type of input mechanism as used presentlyfor CONSTRAINTS, or earlier for the "GLOBAL CONSTRAINTS LIST.Furthermore, perhaps I want to know how many structures havetwo cyclopropanes; or a cyclopropane and a hydroxyl; or, ashappened to me the other day, a hydroxyl and an N-H. Thus,combinations of SURVEY patterns can be useful, using relationaloperators such as the FORTRAN AND, OR, LE , etc. Such a facility

would be particularly useful, I think, in planning futureexperiments. Another potentially useful addition might be for

SURVEY to remember the compound numbers of structures including

the pattern. That would allow the user to DRAW some of thesestructures so as to see how the pattern is incorporated.

REACT is another approach of great potential, but it has to be

expanded so that it can deal with more than a few hundredstructures. My hunch is that REACT could provide an addi-tional means of using the mass spectrum to PRUNE structures.While it is virtually impossible to write a general msinterpreter, one might be able to tailor a limited inter-preter for a given list of candidate structures. If the

I

January 11, 1979Dr. Dennis H. Smith

3

unknown has only one double bond, for example, it might bereasonable to expect that at least the more intense fragmentsresult from allylic cleavages. The difficulty, of course,is that one may have used these major fragments in thesuperatoms. Again, I've little experience with REACT, sopossibly these suggestions are witless.

At present, I have some doubt about STEREO. I have a veryhigh regard for Jim Nourse, and I am sure that his contri-bution is an intellectual tour de force of some magnitude.But, unless someone develops a generalized octant rule, Idon't see how STEREO helps me. to trim down a long list ofstructures. I plan to show the program to Ann Hunt, ourresident "chiropticist, " and we will try it on some of ourcompounds. But at present, I cannot foresee how STEREOwill help.

The workshop, incidentally, was very well run, for which Iam sure you deserve much credit. I probably benefited morethan most participants because I was able to work during theevenings as well. The key to the success of any workshopis having lots of time to work, oddly enough, and that iswhat I had. I must also confess that I very much enjoyedthe crab salads at the Faculty Club.

Sincerely,

SEARCH LABORATORIES

\

Dorman, Ph.D.Research ScientistPhysical Chemistry ResearchDepartment MCS 25

DED.vr

US.DepartmentofENERgy

ARGONNE NATIONAL LABORATORY Nahonal ENERgy Software Center9700 SoutlhCass Avenue, Arqonne, IlliNOis 60459 TelePU)ni

February 22, 1979

E. A. FeigenbaumStanford UniversityComputer Science DepartmentStanford, California 94305

Dear Mr. Feigenbaum,

We have had a request for the DENDRAL program. Would you pleasetell me its status and plans for distribution?

Thank you for your assistance.

National Energy Software Center

cc: B. G. BuchananJ. Lederberg

IMSIS

FEB 26 lb/y

E.A. Ftlbu.ui.u

Tlhe UNivEßsiTy of ClnicAqo ARQONNE UNiVERSiTiES AsSOCiATJON

Sincerely,

Jan Mockler