PUBLICATIONS: D. H. SMITH Page 2

Il.

12.

13.

14.

15.

16.

17.

18.

19.

20.

The Lunar Sample Preliminary Examination Team, "Preliminary Examination
of Lunar Samples from Apollo 12," Science, 167, 1325 (1970).

D. H. Smith, "Mass Spectrometry," Chapter X in Guide to Modern
Methods of Instrumental Analysis, T. M. Gouw, Ed., Wiley-Interscience,
New York, 1972.

 

D. H. Smith, R. W. Olsen, F. C. Walls and A. L. Burlingame, "Real-time
Mass Spectrometry: LOGOS--A Generalized Mass Spectrometry Computer
System for High and Low Resolution, GC/MS and Closed-Loop Applications,"
Anal. Chem., 43, 1796 (1971).

A. L. Burlingame, J. S. Hauser, B. R. Simoneit, D. H. Smith, K.
Biemann, N. Mancuso, R. Murphy, D. A. Flory and M. A. Reynolds,
"Preliminary Organic Analysis of the Apollo 12 Cores," Proceedings of the
Apollo 12 Lunar Science Conference, E. Levinson, Ed., M.1.T. Press,
Cambridge, Mass., 1971, p. 1891.

 

D. H. Smith, "A Compound Classifier Based on Computer Analysis of Low
Resolution Mass Spectral Data," Anal. Chem., 44, 536 (1972).

D. H. Smith and G. Eglinton, "Compound Classification by Computer
Treatment of Low Resolution Mass Spectra-Application to Geochemical and
Environmental Problems," Nature, 235, 325 (1972).

D. H. Smith, N. A. B. Gray, C. T. Pillinger, B. J. Kimble and G.
Eglinton, "Complex Mixture Analysis - Geochemical and Environmental
Applications of a Compound Classifier Based on Computer Analysis of Low
Resolution Mass Spectra," Adv. in Org. Geochem., 1971, p. 249.

 

D. H. Smith, B. G. Buchanan, R. S. Engelmore, A. M. Duffield, A.
Yeo, E. A. Feigenboum, J. Lederberg and C. Djerassi, "Applications of
Artificial Intelligence for Chemical Inference, VIII. An Approach to the
Computer Interpretation of the High Resolution Mass Spectra of Complex
Molecules. Structure Elucidation of Estrogenic Steroids," J. Amer. Chem.

Soc., 94, 5962 (1972).

D. H. Smith, A. M. Duffield and C. Djerassi, "Mass Spectrometry in
Structural and Stereochemical Problems, CCXXII. Delineation of
Competing Fragmentation Pathways of Complex Molecules from a Study of
Metastable lon Transitions of Deuterated Derivatives," Org. Mass.

Spectrom., 7, 367 (1973).

P. Longevialle, D. H. Smith, H. M. Fales, R. J. Highet and A. L.
Burlingame, "High Resolution Mass Spectrometry in Molecular Structure
Studies, V. The Fragmentation of Amaryllis Alkaloids in the Crinine

Series," Org. Mass Spectrom., 7, 401 (1973).

a6

98 ARO ctl a ae
PUBLICATIONS: D. H. SMITH Page 3

21.

22.

23.

24.

B. R. Simoneit, D. H. Smith, G. Eglinton and A. L. Burlingame.

"Applications of Real-time Mass Spectrometric Techniques to Environmental
Organic Geochemistry, If. San Francisco Bay Area Waters," Arch. Env. |

Contam and Tox., 1, 193 (1973).

D. H. Smith, B. G. Buchanan, R. S. Engelmore, H. Adlercreutz and C.

Djerassi, "Applications of Artificial Intelligence for Chemical inference,
IX. Analysis of Mixtures Without Prior Separation as Illustrated for
Estrogens," J. Amer. Chem. Soc., 95, 6078 (1973).

 

D. H. Smith, B. G. Buchanan, W. C. White, E. A. Feigenbaum, J.

Lederberg and C. Djerassi, “Applications of Artificial Intelligence for

Chemical Inference, X. INTSUM - A Data Interpretation and Summary
Program Applied to the Collected Mass Spectra of Estrogenic Steroids,"
Tetrahedron, 29, 3117 (1973).

G. Loew, M. Chadwick and D. H. Smith, "Applications of Molecular
Orbital Theory to the Interpretation of Mass Spectra. Prediction of

Primary Fragmentation Sites in Organic Molecules," Org. Mass Spectrom. ,

7, 1241 (1973).

7
SECTION Il ~ PRIVILEGED COMMUNICATION
BIOGRAPHICAL SKETCH

(Give the fottowing information for all professional personnel listed on page 3, beginning with the Principat Investigator,
Use continuation pages and follow the same general format for sach person}

 

 

 

 

 

 

 

NAME TITLE BIRTHDATE (Ma, Osy, Yr.)
Sridharan, Natesa S. Research Associate 10/2/46
PLACE OF BIRTH (City, State, Country) PRESENT NATIONALITY (/f non-US citizen, SEX
indicate kind of visa and expiration date) .,
India;
Madras, India 5/73-U.S. permanent residence (R) Mate (7) Femate
EDUCATION (Begin with beccelaureate training and include postdoctoral) * -
YEAR SCIENTIFIC
INSTITUTION AND LOCATION DEGREE CONFEARED FIELD
Indian Institute of Technology, Madras Bachelor of
India Technology 1967 Electrical Engineering
State University of New York, Stony Brook |M.S. 1969 Computer Science
Ph.D. 1971 Computer Science

 

 

 

 

HONORS
University Fellow - 1968-1971, SUNY Stony Brook; Graduate Assistant - 1967-1968, SUNY

Stony Brook; Siemens! Award (awarded for top rank in Electrical Engineering) - 1967, ITT
Madras; National Merit Scholarship - 1963-1967, ITT Madras

MAJOR RESEARCH INTEREST ROLE IN PROPOSED PROJECT

Computer Application in Chemistry
and Medicine Research Associate

 

RESEARCH SUPPORT (See instructions}

RESEARCH AND/OR PROFESSIONAL E XPERIENCE (Starting with present position, dist training and experience relevant to sraa of projpct List aif

or most representative publications, Do not exceed 3 pages for each individual.)
1971-present Research Associate, Heuristic Programming Project, Stanford University
1970-1971 Consultant, IAC Computer Corp., Long Island, N.Y.

Sridharan, N.S., "An Application of Artificial Intelligence to Organic Chemical Synthesis"
Doctoral Thesis, State University of New York at StonyBrook, 1971.

Sridharan, N.S., "Search Strategies of Organic Chemical Synthesis", Third Internationa]
Joint Conference on Artificial Intelligence (3IJCAI), Stanford, 1973

Sridharan, N.S. (co-author), "Heuristic DENDRAL: Analysis of Molecular Structure", Proc.
NATO Advanced Study Institute, Amsterdam, 1973.

Sridharan, N.S. (co-author), "Heuristic Theory Formation'', Machine Intelligence, Volume 7,
Edinburgh, 1972.

 

Hin 398 (FORMERLY PHS 390)

Rav. 1/73 ;
U, 8, GOVERNMENT PRINTING GFFICE : 1871 0-45 .-456

 
  
 
SECTION II — PRIVILEGED COMMUNICATION
BIOGRAPHICAL SKETCH

(Give the folowing information for all professional personne! listed on page 3, beginning with the Principal Investigator.
Use continuation pages and follow the same general format for each person.)

 

 

 

 

 

 

 

 

 

 

 

 

NAME TITLE BIRTHDATE (Ma,, Day, Yr.)
Brown, Harold D. Associate Professor July 12,1934
PLACE OF BIRTH (City, State, Country} PRESENT NATIONALITY (/f non-U.S. citizen, SEX
, indicate kind of visa and expiration date) ;
South Bend, Indiana U.S. . (imate Cl Femate
EDUCATION (Begin with baccalaureate training and include postdoctoral)
YEAR SCIENTIFIC
INSTITUTION AND LOCATION OEGREE CONFERRED FIELO
University of Notre Dame, Notre Dame, M.Sc. 1963 Mathematics
Indiana
Ohio State University, Columbus, Ohio Ph.D. 1966 Mathamatics
(No Baccalaureate Degree)
HONORS
Summa Cum Laude - Notre Dame
MAJOR RESEARCH INTEREST JROLE IN PROPOSED PROJECT
Applied Discrete Mathematics - Computer
Science | Research Associate

 

RESEARCH SUPPORT (See instructions}
Principal Investigator, NSF-GP-16793 (Expires March, 1974)

Pending Proposal NSF (Proposed starting date September, 1974)

 

RESEARCH AND/OR PROFESSIONAL EXPERIENCE (Starting with present position, dist training and experience refevant to erea cf project List aff
or most representative publications, Do not exceed 3 peges for each individual.)

Visiting Associate Professor, Computer Science, Stanford University , 1971-72, 1973-present
Associate Professor, Mathematics, Ohio State University, 1966-

Visiting Professor, Mathematics, Rhine,Westf. Tech. Hoch., Aachen, 1972 and 1973

Visiting Member, Courant Institute, New York University, 1967-68

Instructor/Assistant Professor, Assistant Chairman, Mathematics, Ohio State U., 1963-65
Assistant to the Chairman, Mathematics, University of Notre Dame, 1960-63

Director or Associate Director, NSF-SSTP, 1964-70

 

NIH 398 (FORMERLY PHS 396)
Rev, 1
* m3 U.S, COMERYMENT PRINTING OFFICE : 1077 C €59- 78 A

ee AE TE ON ETRE
Vitae Page 2

Publications
Near Algebras, Ill. J. Math, 12(1968), Pg. 215.

Distributor Theory in Near Algebras, Comm, Pure App. Math.
XX1(1968), Pee 5356

An Algorithm for the Determination of Space Groups, Math Comp.
23(1969), Pe. 499.

Some Empirical Observations on Primitive Roots, with H. Zassenhaus,
J. Number Theory 3(1971), Pe. 306.

A Generalization of Farey Sequences, with K, Mahler, J. Number
Theory 3(1971), pe. 364,

Basic Computations for Orders, Stanford CS Memo STAN~CS-72-208,

An Application of Zassenhaus' Unit Theorem, Acta Arith. XX(1972),
Pe. 154.

Integral Groups I: The Reducible case, with J. Neubiser and
H. Zassenhaus, Numer, Math. 19(1972), Pg. 386.

Integral Groups II: The Irreducible Case, with J. Neubl’ser and
H. Zassenhaus, Numer, Math. 20(1972), Pg. 22.

’ Integral Groups III: Normalizers, with J. Neubuser and
H, Zassenhaus, Heth. Comp. 27(1973), Pee 167.

Constructive Graph Labeling Via Double Cosets, with L. Hjelmeland
and L, Masinter, Discrete Math, in press and Stanford CS Meno
STAN~CS-72-318,

An Algorithm for the Construction of the Graphs of Organic Molecules,
with L, Masinter, Discrete Math. in press and Stanford CS hemo
STAN-CS-73-261,. |

The Crystallographic Groups of 4-dimensional Space, with J. Neubiser,
H. Wondratschek and H, Zassenhaus, Wiley-Interscience in press.

JO
SECTION Il — PRIVILEGED COMMUNICATION
BIOGRAPHICAL SKETCH

{Give the foliowing information for all professional personnel listed on page 3, beginning with the Principal Investigator.
Use continuation pages and fotlow the same general format for each person.]}

 

 

NAME TITLE BIRTHDATE (Ma, Day, Yr)
DROMEY, Robert Geoffrey Research Associate 11/21/46
PLACE OF BIRTH (City, State, Country) PRESENT NATIONALITY (/f non-U.S citizen, SEX

indicate kind of visa and expiration date)

 

 

Castlemaine, Victoria, Australia ‘Australian, J-1 Visa, Exp. 10/8/74 () Mate Cl Femate
EDUCATION (Begin with baccalaureate training and include postdoctoral)

 

 

 

YEAR SCIENTIFIC
INSTITUTION AND LOCATION DEGREE CONFERRED FIELO
Swinburne College of Technology, Diploma of 1968 Chemistry
Melbourne, Australia Appl. Chem.
La Trobe University Ph.D. 1973 Molecular Science
Melbourne, Australia

 

 

 

 

HONORS csTRO Postdoctoral Studentship

Commonwealth Postgraduate Research Scholarship

Walter Lindrum Memorial Scholarship

Equivalent of First Class Honors Master of Science Preliminary (1969)
MAJOR RESEARCH INTEREST anniication of ROLE IN PROPOSED PROJECT

Artificial Intelligence Techniques to Bio;
Medical and Chemical Problems. Research Associate
RESEARCH SUPPORT (See instructions)

 

 

RESEARCH AND/OR PROFESSIONAL E XPERIENCE (Starting with present positicn, list reining and experience refevent to ares of project List ail
or most representative publications, Do not exceed 3 pages for each individual.)

1973 DENDRAL Project, Stanford University, Computer Science Department
1973 Software Development for Graphics Systems, LaTrobe University, Computer Centre
1969-73 Construction, development and applications of an on-line photoelectron spectrometer

LaTrobe University, Chemistry Department
1969-73 Application of Deconvolution Techniques to the Processing of Experimental Data.

Publications:
"Deconvolution and Its Applicatim to the Processing of Experimental Data", Intl. Journal of

Mass Spectrometry and Ion Physics, 1970, 4. (co-author).

"Inverse Convolution in Mass Spectrometry", Intl. Jnl. Mass Spec. Ion Phys.,1971, 6. (co-
author).

"A Combined Time Averaging-Deconvolution Technique Applied to Electron Impact lonization
Efficiency Curves", Internation Journal of Mass Spectrometry & Ion Physics, 1971, 6.
(co-author).

"The Perfect Direction and Velocity Focus at 254934' in a Cylindrical Electrostatic
Field'', Reviews of Scientific Instruments, 1973, 44. (co-author).

 

MiH 398 (FORMERLY PHS 396)
Rev, }
" 3 U, S. GOVERNMENT PRINTING CRFICE . 0971 O - wot

PEE Ts F
R. G. Dromey

"Detection of Spin-Orbit Splitting in the Photoelectron Spectrum of Oot by Deconvolution",
Chem. Physics Letters (in press), 1973. (co-author)

"The Effect of Finite Line Widths on the Interpretation of Photoelectron Spectra", Journal
of Electron Spectroscopic (accepted for publication). (co-author).

"An On-line Ultraviolet Photoelectron Spectrometer for High-Resolution Studies of Molecular
Structure'’, Australian Journal of Chemistry (accepted for publication). (co-author).

"Photoelectron Spectroscopic Correlation of the Molecular Orbitals of the Alkanes and
Alkyliodides", Journal of Molecular Structure (submitted for publication). (coauthor).

“Comparison of the Photoelectron Spectra and the Photoionization Efficiency Curves for
the Alkyliodides", Transactions of the Faraday Society (submitted for publication).
(co-author).

"A Convolution-Deconvolution Algorithm Using Fast Fourier Transforms), Decuscope,
1973 (in press).

Tar
RESEARCH PLAN
RTOMOLECULAR CHARACTEPRTZATION: ARTIFICIAL INTELLIGENCE
A Program of Resource-felated Research
T. INTRODUCTION
A. %Ibijectives

B. Background and Rationale
C. Relationship to AIM-SUMFX and the Genetics Research Center

It. SPECIPIC AIMS

Itt. MFTHODS

Tv. SIGNTFICANCE OF PROPOSED RPESEARCE
V. FACILITIES & EQUIPMENT

Vr. ET BLIOGRAPHY

Table 1

Pigures 1-3

Appendix A: Letters of Interest
Appendix RB: 1973 Annual Report to the NTH

S$
I. INTRODUCTION

This renewal application is intended to. sustain and augment the
capabilities of the mass spectrometry (MS) program which has
served as a major institutional resource at Stanford for some
yoars. With previous support from NASA and NSF it has made
possible a highly interdisciplinary set of research projects
ranging over: artificial intelligence (AT) in bionolecular
tharacterization, natural product chemistry, clinical biochemical
studies on steroids, and the machanisms of molecular fragment
Formation in mass spectrometry. While the facility equipment for
mass spectrometry has been funied mostly by other agencies,
connected research programs embrace several NIY research projects
as wall. In addition, this activity was closely coupled with the
ACME Madical School computer resource (1966-1973) and will have
Similar associations with the new ATM-SUMEX conputer resource
cetently fanded by the BRR (see Section T.7).

Previous support reflects the diversified facets of this
interdisciplinary research. NASA haS supported projects in new
iasteouneatation, including the initial mass spectrometer-computer
link, NSF has supported chemical research, and ARPA has supported
auc artificial intelligence research and initial application to
mass spectrometry, Overall cuthacks have forced NASA to reduce
Funding for this area of research despite their interest. Under
ARPA support to Drs. Feigerbaum and Lederberg for AI research, the
DENDRAL program became recognized as one of the most successful AI
applications programs. However, ARPA is chartered to fund
frontier computer science research and no longer provides funds
for the DENDRAL applications programs. ARPA has indicated a
reluctance to continue funding to this groun for the theory
formation work in chemistry, although we expect to continue to
ceceiv2 ARPA support for more theoretical aspects of our research
oroaram (@.g9., automatic programming).

We previously submitted a comprehensive proposal to the NIH
(R8-00785, 3/28/73) which included an application for the
AIM-SUMEX computing resource and a renewal of the existing DENDRAL
grant {RR-00612). This proposal was approved for 5 years by the
National Advisory Research Resources Council. Certain
reservations were, however, communicated to us: they concerned
aspecially what we must agree was an anbitious effort to close the
sontrol loop for "Yintelliqent automation" whose costs overreached
the immediate utility of the expected result. During subsequent
discussions with the Biotechnology Resources Branch, taking into
account the council review and a number of diverse policy issues,
we ayreed administratively to seqment the two components of the
yriginal proposal. The AIM-SUMEX portion of the original proposal
(excluding DENDRAL) was recently funded for 5 years as a national
resource for artificial intelligence in medicine. The present
proposal for resource-related research in biomolecular
characterization and artificial intelligence is an elaboration of
thea DENPRAL portion incorporating intensive reexamination and
revision of the previous proposal.

Bith the dALfferentiation of priorities represented by AIMN-SUMEX,
the Sanetizs Research Center (37C), and continuing work on
artificial intelligence under Dr. Feigenbaum's leadership, the
asreseit renewal application places more emphasis than heretofore
an raal-world oriented applications. Correspondingly, we have
aqceel that it is now more appropriate that Dr. Djerassi should be
Aesignated as Principal Investigator in this phase of our work.

Rs outline? in section B.2, the interests and responsibilities of
Professors Djerassi (Chemistry), Feigenbaum (Computer Science) and
Laderberg (Genetics) have been closely interdigitated. With their
further rtonnections with many colleagues, these programs enjoy a
high deqree of university-wide participation. For example, the
Janatics Department is also closely affiliated with Biology,
Biochemistry, Pediatrics, Psychiatry and Medicine through joint

appointments or joint research projects or both.

This breadth

would be difficult to obtain except at a few institutions where
the medical school is both academically ani geographically
inteqrated with the university to the degree that characterizes
the Stanford University environment.

GLOSSARY OF ABBREVIATIONS

ACME Advanced Computer for Medical Research (Nih-funded computer
resource, 1968-1973)

AT artificial intelligence

AIM-S'MEX- A comprehensive computer resource intended to serve
the national requirement for artificial intelligence
in medicine. This will he implenented at the Stanford
University facility called AIM-S'MEX

AROA Advanced Research Projects Agency of the Department of
Defense,

BR Biotechnology Resources Branch

T32MR carhon-12 magnetic resonance

3¢ gas chromatography or gas chromatograph

3R2r Genetics Research Center (Stanford, J. Lederberg,
Principal Investigator; NIGMS-approved and
awaiting funding. Grant #P01-GM 20832-01)

HERES high resolution mass spectrometry

TR infra-red

IRL Tnstrumentation Research Laboratory
(Stanford Genetics Department)

Lees low cesolation mass Spectrometry

¥CD magnetic circular dichroisn

MS Mass Spectrometry or mass spectrometer

VASA National Aeronautics & Space Administration

yep nuclear magnetic resonance

NSF National Science Foundation

IRD optical rotatory dispersion

DULSACME a modified version of the PL-1 computer language (for the
Stanford ACMF computer facility)

SUMEX Stanford University Medical Experimantal Computer Resource
(NIH funied computer resource, 1973-1978)

Ty ultra-violet

JE
A. OBJECTIVES:

Core Research.
The funds now applied for would permit

1) the continued funding of the 4S laboratory as a biomolecular
sharactterization resource;

2) advanc2ment of laboratory instrumentation capability in
specific areas of GC-HRMS and the exploitation of metastable peak
analysis.

3) the fucther development of AI computer techniques to match the
instrunentation. This work will emphasize practical utilization
for applications in biomolecular characterization connected with
other on-qoing biomedical research programs. ft will include, for
2xanple, a) the analysis of mixtures by GT/MS; b) metastable peak
analysis for difficult problems of pure compounds and of mixtures
not rceadiily separable by GC; c) optimized data analysis for
sharacterization of MS peaks ani d) heuristic analysis of spectra
for the molecnlar ion composition.

Juc projact is the only systematic effort, to our knowledge,
currentiy underway in this country for computer assisted structure
slicidation. Subsequent to our early publications, an intensive
program has been mounted in Japan in similar areas. This
situation may be contrasted with computer assisted organic
synthesis, an area receiving considerable attention from several
research groups. These capabilities can be beneficially provided
to a wider community via the ATM-SUMEX resource. Research on the
amulation of human intellect by computer programs will undoubtedly
influance the efficiency with which chemical research can be
applied to ever more complex problems of health, e.g.,
intermediary metabolism and its pathologies; environmental
influenctes on health; the development and critical validation of
new therapeutic agents.

The athievament of these objectives depends on the continued
naintenance and development of the DENDRAL AT programming system
(sae balow). The advent of tha AIM-SUMEX facility will remove
some of the serious computational limits on the exercise of this
system that have delayed recent prograss.

PFducation.

Tn our university setting, pre-doctoral and post-doctoral
aducation of course constitutes a part of our mission. As far as
is practically possible, research participation in the DENDRAL
program has been coupled with dissertation work by graduate
students and post-doctoral research experience respectively.
Examplas of people (and their research area) whose education has
been enhansed in this way are the following:

Scadnate Students: J. Simek, pedagogical aspects of the structure
qenerator; Wai Lee Tan, synthesis of new estrogen compounds; H.
PFqgert, 13°MR of amines and steroidal ketones; C. Van Antwerp,
13CMR of steroidal alcohols; c. Farrell, theory formation fron
mass Spectral data: L. Masinter, development of the structure
qJenerator: M. Stefik, AT applications to chemistry.

37
Postdoctoral Fellows: G. Dromey, theory formation from analytical
Jata: &. Gritter, mass spectral fragmentation of hiologically
active steroids; 8&8. Carhart, analysis of 13CMR spectra by
DENDRAL-like programs; $. Hammerum, development of better
Fragmentation rules for progesterones.

Formal organization.

Phis project has been a long-term commitment of Djerassi,
Lederberg and Feigenbaum functioning in effect as
so-investigators. We coordinate our activities with day-to-day
tontacts ia the pursuit of convergent research objectives. In the
light of the extension of our collaborativ2 activity during the
last t#o y32ars, we are now organizing a formal advisory group to
jaclude, in addition to ourselves, H. Cann, J. Barchas, and E£. Van
Tamelen. This group will advise the principal investigators on
the direction of the program with respect to allocating available
facilities and seeking out and helping other collaborators. This
Jesignuation simply recognizes the fact that many of our colleagues
have alreaiy heen engaged in r2lavant collaborative research with
4S. A MS resource has recently been funded at the University of
Talifornia/Berkeley, under the direction of Dr. A.L. Burlingame.
Drs. Djerassi and Burlingame have recently engaged in some
tollaborative research which was made more successful by the
sharing of facilities ard expertise available at one institution
but not at the other. We would hope to maintain and strengthen
these contacts to avoid unnecessary duplication of effort.

We plan to discuss with Dr. A.L. Burlingame the most appropriate
procedures for coordinating the related activities of our
respective programs at the University of California/Berkeley and
hara. Phis may take the form of raciprocal membership in advisory
rommittees.

The “hardware resource" to which this application is pegged has
n2aen identified as the MS facility. While these instruments alone
represent an investment of over $300,090, funded previously by
several agancies, they do not cepresent th2 most ilaportant resource.
He would uses this designation instead for the working team led by
the princioal and co-investigators. The skills embraced by this
ycoup incliite, as mentioned, computer science, structural organic
chemistry, molecular biology, instrumentation engineering anda
wide cange of other disciplines. They are represented not only in
the princinal professors but in a diversified and accomplished
professional research staff {see Budget Justification). The
program for which funds are now requested is the vital means by
which the interests of this group can be sustained ina
corrdinated effort that would be very costly both in funds and in
tine if it had to be reconstructed from scratch. Without the
finantial support now requested, this line of collaborative
research will have to be abandoned, with it a unique style of
interdisciplinary collaboration, and the MS facility will be
terminated.

Se
mR, BATKGROUND AND RATIONALE
1. The Structure Flucidation Problen

a) The General Problem. Analysis of molecular
structure is a major activity in our program of resource related
research. For the specific task of elucidating molecular
structures, i.e., the topology of atom-to-atom connectivities,
analysts utilize a mixture of information derived from chemical
proceiures and spectroscopic techniques. Each item of
information, if not redundant or uninterpretable, contributes to
the solution of the problem. Chemists draw upon a tremendous body
af specific knowledge about the task area (e.q., clinical
shemistry, biochemistry), molecular structure, spectroscopic
techniques, etc., in order to piece together this information and
iafer the structure of molecules. These features, and the
relative simplicity of the final concept of a structure, make the
prohlem particularly well-suited for applications of the
techniques of AI to assist research workers performing the task.

b) Njerassi'ts Laboratory. Professor Djerassi has been concerned
with structure elucidation problems since the beginning of his
shamical research. His activities at Stanford have been concerned
heavily with the application of particular spectroscopic
tachniques to structural studies of bismedically important
tonpounds. These techniques include optical rotatory dispersion
(ORD) and, more recently, maqnetic circular dichroism (MCD) (both
of them supported initially by the NIH). Since 1961 he and his
yroup have also been concerned with MS because of the power of the
technique, in terms of specificity and sensitivity, as an
analytical tool for structure elucidation. Four books and
approximately 250 articles on 4S have been puhlished by him and
4is colleayjues.

The technique of MS does not suffice for all structure
Jatarmination problems, but it is a very powerful tool in areas
where there exists a body of knowledge about the MS behavior of
related molecules. When sample size is limited MS may well be the
only technique that can be utilized. The recent availability of
hiqh resolution mass spectrometers has mad2 HPMS the technique of
choicte for many applications because under ideal conditions the
axact mass number uniquely specifies the the empirical formula of
a molecule or fragment. On a parallel course, the technique of
37/™S, routinely available with low resolution mass spectrometers
(GC/LEMS), has revolutionized investigations wherever complex
mixtures ace encountered. All of the above considerations argue
that an extension of MS at Stanford to provide routine GC/LRMS and
SC/HRMS analyses would be the next logical step to assist
researchers depending on this facility for solutions of their
structure elucidation problems,

2. Historical Background

a) Mass Spectrometry Lahoratory. Prior to the existing DENDRAL
qrant, the groundwork was laid for computerization of the existing
mass spectrometers, an Associated Flectrical Industries MS-9 high
resolution mass spectrometer and an Atlas TH-4 low resolution mass
spectrometar. This work, supported primarily by NASA via the

37
Trastrunmontation Research Laboratory (T&L) in the Department of
Sanetics, cesulted in link-up to the then axisting ACME computer
facility via a PDP-11 mini-computer which acted as a buffer
between the spectrometers and ACME. Initial data acquisition and
reduztion programs were written for the system and utilized ona
Limited basis. The funding of the DENDRAL proposal, NIH grant
RPR-612 (May 1,1971-present) in conjunction with additional
resourztes provided by the IRL resulted in a najor improvement to
thes2 capabilities. The fruits of these efforts are described
uniter section I.B.3 {below}.

b) Summary of Early DENDRAL Development.
In 1964, Lederberg devised a notational algorithm for chemical
structures (termed DENDRAL) that allowed questions of molecular
structure to ba framed in precise graph-theoretic terms. (Refs.
1,3-5,12). He also showed how to use the DENDRAL algorithm to
generate complete and irredundant lists of structural isomers.
(Refs. 1,5).

In 1965-66 Lederberg and Feigenbaum began 2axploring the idea of
using the isomer generator in an artificial intelligence program -
searching the space of possible structures for plausible solutions
to a problem much as a chess-playing program searches the space of
leqal moves for the best moves. (Refs. 7,12). This approach
quaranteas that every possible solition to a problem is considered
- aither inolicitly, as when whole classes of unstable structures
ace rajected, or explicitly, as when complete molecules are tested
for plausibility. In either case, an investigator easily
jJetermines the criteria for rejection and acceptance and knows
that no possibilities have heen forgotten. This approach also
quarantees that structures appear in the list only once - that
autonzorphic representations of the samo complex molecule have not
b22n included. In both these respects the computer program has an
advantage over manual approaches to structure elucidation.

c) Tnitial collaboration with Djerassi. (Refs. 14,15,19,
20, 21,22,24).
Lederberg and Feigenbaum realized that (a) only through
application to real problems could the AI approach be materially
advanced and critically evaluated, and (b) MS appeared to be a
fruitful applications area. MS appeared to be an excellent
problem area because of the close relationship between spectral
Fraqmentation patterns and molecular structure for many classes of
noleculas. Dijerassi'’s interest and expertise - and daily
interactioa between members of his group and the AI group - led to
a series of joint publications describing the approach and initial
results of the programs. The success of these collaborative
afforts led to the proposal to the NIH for initial funding to
extend these efforts.

4) Efforts Under NIH Funiing for DENDRAL. (Refs. 25-41).
The initial funding by NIH provided the opportunity to upgrade the
instrumentation and computer programs. In particular we were able
to mount a concerted project on both the analysis of mass spectra
of bionedically important compounds and the mathematical aspects
of molecular structure. Progress reports to the NIH describe this
research in detail. The most recent annual report appears in
Appendix B. A series of publications directed to audiences both
in computer science and chemistry are listed in the bibliography.
The following section (Section 3) summarizes the capabilities for

40
strusture alucidation which, in thamselves, constitute an
important result of past work.

a) Related Research.
An important side effect of tha DENDRAL project is the extent to
which additional research was inspired and carried out to fill
qyaps in existing knowledge. This research, not supported by the
QNENDAAL grant, has been beneficial to on-going DENDRAL work, and
vice-versa. Publications which have arisen from this research are
Listed in the bibliography (Refs. 58-70). A brief review of these
publications should indicate the need for precise specification of
the kaowleige elicited from chemists and used in computer
programs. AS an example, consider the description and application
yf an early algorithm for generation of cyclic structural isomers
(21). This paper considered the problem of spectroscopic
Jiffarentiation of isomers of [°6H100. Unsaturated ethers fall in
one of the classes of isomeric compounds which must be considered,
but the MS of unsaturated ethers had not been investigated
Systematically. This work was subseguently carried out in
Professor Djerassi's laboratory independently of DENDRAL support,
but of benafit to DENDRAL (62). Other examples will be found in
the Bibliography (Refs. 58-79).

3. Existing Capabilities

#o have worked to develop distinctive capabilities for molecular
structure elucidation, bringirg together a high quality HRMS
3ysten and AI programs applied to biomolecular characterization.
The feasibility of our analytical approach has been demonstrated
in saveral problem areas, basel upon the development both of a MS
syst2n and a general set of computer programs for use in new
areas.

The princival capabilities are summarized below. These are now in
yeiag and were developed primarily under NIH funding to this
project, with additional support supplied by ARPA and NASA in
specific areas. (These agencies have reduced funding levels for
this work because overall cutbacks have forced NASA to cut out
this area of research despite their interest and ARPA is chartered
to provide funds for frontier computer science research but not
for applications. Thus the NIH is the principal of support for
Future development of anplications programs in the
interdisciplinary area of artificial intelligence/heaith related
chemical problems.)

a. HRMS System and Coupled SC/LRMS System.
We have coupled the NIH-supported Varian-MAT 711 High Resolution
Yass Spectrometer with a Hewlett Packard Gas Chromatograph and
Jeanoastrated its utility for GC/LRMS analysis of such difficult
analytic problems as the free sterols (i.e., not derivatized)
isolated from marine and other sources. Advanced data reduction
techniques for this instrument were written for use with the ACME
conputer system (360/50) and row exist in Stanford's new 370/158
which tontinues to support the PL/YACME language. SC/HRMS scans on
extracts from urine and amniotic fluid demonstrated this system's
ztapability to provide high quality mass measurements on complex
nixtures obtained from biological sources. An example of one
SC/HRMS run on the amino acid fraction of amniotic fluid is
presented below (Sec. III.D).

4/
b. DENDRAL Structure Senerator (Refs. 1-6,14,31,37, 38, 40,41)
Tho DENDRAL Structure Generator progran accomplishes exhaustive
and irredundant generation of isomers, with and without rings.
This proqram quarantees consideration of every candidate structure
- either implicitly, as when whol® classes of structures are
forbidjten, or explicitly, as when individual compounds in a class
are specified. It corresponds to the "legal move generator" of
convouterized chess playing and other heuristic programs.

c. DRENDRAL Planner (Refs. 25, 28,33)
qe have written a very general set of computer programs for
Aaterminingy structural features from analytical data in
well-defined areas. Such general planning programs have been
written fot low and high resolution MS, interpreted proton NMR
spectroscopy and 13CMR data.

J. INTSUM (Refs. 26,29,34,35)
INTSIM is a computer program that aids in finding interpretive
rules for “4S. The program interprets a large collection of MS
Jata actoriing to criteria specified by the investigator. Then it
summarizes the data to show which of the possible interpretations

sean most plausible.

@. PULEGEN {Refs. 26,35)
ROLEGEN is the current rule generation program that suggests
various rules of interpretation for the MS data summarized by
INTSIM. Although not finished, the program can provide useful
assistance in practical theory formation.

f. Ancillary Techniques
1. The MS facility provides other types of experiments in MS,
including ultra-high resolution measurements (masses determined
via peak matching), defocussed metastable ion determinations
(Bacbar-Flliott technique) and low ionizing voltage experiments.
These data are utilized by both scientists and programs where
appropriate.
2. Additional computer prograns provide added problenm-
solving assistance.

a. Predictor program for predicting major features of mass spectra.

b. Programs for drawing and displaying chemical structures.

c. Subroutines developed in conjunction with or existing as parts of
the Structure Ganerator for problems of partitioning, construction
»f vertex-graphs, and constructive graph labelling. These can he
applied? to answer certain questions of isomerism which do not
reyuire the complete generator. For example, the labelling
algorithm can list all structures resulting from substituting
sites of a carbocyclic skeleton with stated numbers of different
Functional groups.

g. Other Spectroscopic Techniques
Available to us are the facilities of Professor Djerassi's
laboratory for work requiring additional spectroscopic data. Also
available on a fee for service basis are extensive spectroscopic
facilities (NMR, I.8., and U.V.) of the chemistry department.
These woulld be utilized for collecting additional data on
particular structure problems and gathering data on known
tompounds {particularly in the area of 13CMR) as the AI programs
beacoma knowledgable about other spectroscopic information.

Fr
h. Chemical Facilities
The staff and facilities of the chemistry department represent
substantial synthesis capabilities and general chemical know-how.
This resource can be called upon to provid? assistance in
synthesis of model or labelled compounds, derivatization of
mixtures, and so forth. For example, a graduate student in
shemistry is presently engaged in thesis research dealing with the
laboratory synthesis of a new astrogen metabolite strongly
suspected to be a component of certain preqnancy urines. The
previously proposed structure of this tompound was one of the
sandidate structures inferred by the planner in a study of
astrogen mixtures (11-dehydroestradiol-17-alpha, ref. 33).

4. User Community
®conogmic utilization of existing and proposed facilities can be
realized by sharing them with a community of users. Lacking
supplementary funds that would be needed for a comprehensive,
naior service facility, this community will include the following
yr>duos, but will be informally available to others.
A. Stanford Community
i) Stanford Chemistry Department (except for Hodgson, all
are heavily supportei by the NIH in their research efforts)
Latters of interest are attached to the proposal
in Appendix A.

Prof. C. Djerassi - Steroids, marine sterols
Prof. W. Johnson - steroids
Prof. E. Van Tamelen - steroids, triterpenoids, other

natural products
Prof. H. Mosher - natural products {(e.g., marine toxins)
Prof. K. Yodqson - biological ligands, ligand-metal complexes
Prof. J. Collman ~ cytochrome P450 models

ii) Stanford Medical School Collaborators

The following research projects in the Stanford
Biomedical Community will furnish samples for mass
spactrometric analysis under the present proposal.
Attached to this proposal (Appendix A) are copies of
the letters of interest in the proposed facility
received from the principal investigators of these
qrants.

Pe. James RF. Trudell, Department of Anesthesia,

Stanford University School of Medicine. Drug

metabolite identification in humans.

Dr. Irene S. Forrest, Biomedical Research Laboratory,
Veterans Administration Hospital, Palo Alto. Drug
metabolite identification in humans.

Dr. I. Rabinowitz and D.I. Wilkinson, Department
of Dermatology, Stanford University School of ,
Medicine. Prostaglandins.

Prof. Fugene D. Robin, Department of Respiratory Medicine,
Stanford UWriversity School of Medicine, Ratio of
NADt/NADH in cells by measnring ratio of oxidized

to reduced redox pairs.

Dr. Leo E. Hollister, Veterans Administration

43
Hospital/sNepartment of Medicine, Stanford University
School of Medicines. Metabolism of Marihuana.

Dr. Hiram 4. Sera, Pharmacy Devartment, Stanford
University Hospital. Drug Identification.

Dr. Sumner M. Kalman, Department of Pharmacology,
Stanford University School of Medicine. Drug and
drug metabolite identification.

Dr. Jack Barchas, Department of Psychiatry, Stanford
University School of Medicine. Neurotransmitters
and. related compounds in man.

De. Keith A. Kyenvolden, chemical Fvolution Branch,
NASA Ames Research Center, Mountain View, Calif.
Amino acids, acids in geochemical samples, structure
of products formed from electrical discharges in gas
mixtures.

Dr. William PR. Fair, Department of Urology, Stanford
University School of Medicine. Identification of
the prostatic antibacterial factor; polyamines
(putrescine, spernine, spermidine) in body fluids

of patients with prostatic carcinoma.

Besiias the user projects just summarized, other major prospects
are in sight. At the time of writing, the chair of pharmacology
is vacant. Conversations with the leading candidate have
indicated a deep-seated interest in GC/HRMS as the principal
analytical tool for broad ranging studies of drug metabolism in
nai.

8. Extramural Users
The davelopment of the techniques of ORD, MS and MCD at Stanford
has beer paralleled with extensive sharing of these resources
nation- ani world-wide in collaborative research efforts, without
any additional funding. Rather than provide routine service,
axperience has shown that discretionary selection of problems
results in better utilization of our peopl? and instrumentation
cesonrces. We would extend this provision of services including
available computer programs, to a limited number of extramural
users. Note, for example, our successful collaboration with
Profassor Adlercreutz, Meilahti Hospital, University of Helsinki,
ar tha identification of estrogens fron body fluids utilizing the
AT planning program {ref. 33).

44
c Relationship to AIM-SUMEY and the Genetics fesearch Center

we

Tha present application is strengthenel by two research projects
related to, but not overlapping, the proposed research of this
grant.

1) AIM-SUWEX (NIH BR-00785, Oct. 1, 1973, thru July 31, 1978,
Principal Investigator, J. Lederberg). This is a resource grant.
to establish a national facility for applications of artificial
intelligence in medicine (AIM). Our own use of this facility will
inctluia SUMEXY PDP-10 computer time and file storage necessary to
run the DENDRAL artificial intelligence programs. This support
will be furnished without charge to the present proposal. It
tepresents an annual investment of about $190,000 in computer time
2yuivalent value.

The ATM-SUMBEX computing facility is shared equally between a
national user community {AIM) and a Stanford Medical School
tommunity. The DENDRAL research will be supported out of the
Stanford portion. The AIM service will be administered under the
9olicy control of a national advisory committee and will be
imolemanted over a national computer natwork. AIM-SUMEX provides
the means for members of the national user community interested in
structure elucidation to access the DENDRAL programs.

2) Genetics Research Center (NIH PO1-SM 20832-01 - approved by
the NISMS Touncil, awaiting funding, Principal Investigator, J.
Lederberg). This research proposal is a comprehensive grant which
would snpport interdepartmental research at the Stanford Medical
3chool in Yedical Genetics, Pediatrics and other clinical
apolications. A section of that proposal concerns the use of
SC/LRMS for screening body fluids for evidence of inborn errors of
metabolism. (This project grew out of the initial DENDRAL grant,
one of the research qoals of which was the analysis of body fluids
using SC/MS). This research on inborn metabolic errors will be
zsonducted jointly in the Stanford Departments of Genetics and
Pediatrics using existing equipment {Finnigan 1015 Quadrupole mass
spectrometar, Varian Aerograph GC and a PDP-11/20 based data

system).

Wo appreciated the value of GC/HRMS analyses of selected extracts
af body flaids (i.e., those containing metabolites not identified
yv routine GC/LRMS data) when formulating the Genetics Research
Teater proposal. Accordingly, a small amount of funding was there
caqguestal For recording selected GI/HRMS data on the GC/Varian MAT
711 mass spectrometer in the Dapartment of Chemistry. If these
funds are awarded, we will negotiate with NIH a suitable
alimination of this minor overlap with the present budget.

YS
ty. SPScrFTte AIMS

Th2 specific aims enumerated in this section will be pursued in
tha highly inter-disciplinary manner that has characterized the
DENDRAL project from the start of its NIH support. The aims are
not disjoiat,but interactive and inter-dependent. For example,
the power of MS and, potentially, other spectroscopic techniques,
tan be anhanced by the use of computer programs to perform various
asnects of structure elucidation and theory formation. From the
starnipoint of computer science, one measure of the utility of
techniques of artificial intelligence is how well they perform in
real-world applications. It is necessary in the development of
these programs to have a source of data and informed, involved
tean-mates able to criticize m2thods and results. The aims are
alabosrated in the methods section.

We have attempted to keep the proposal to a readable length.
Therefore, some detail has been omitted. However, many details
can ba found in the biliography ani we are prepared to provide
additional information Juring the site visit.

1. Enhance the power of the MS resource.

The axisting MS resource, together with computer programs which
axist oar which are proposed (see Aim 2, below), is capable of
solving sone of the structure @lucidation problems of the user
community given computer support for data collection and
reduction. We refer specifically to the areas of GC/LRMS and
roatine, batch HRMS samples, We believe that many of the problems
of the user community require nore powerful technigues (see
Section IIlt). These techniques, specifically GC/HREMS and
Seni-antomatic metastable defocussing, can be provided with a
minimum of cost and effort, thus enhancing considerably the
capzhbilities of the resource.

Jur first aim is to provide the resource with adequate computer
support (replacing the previous ACME system) to enable collection
and reduction of mass spectral data including low and high
resolution scans and data or defocussed metastable ions.

#2 oropose to develop this computer support in the ways described
agelow. {these aims are written to include the work necessary to
imolement the extended PDP-11/20 computer system. A description
of the rationale for this choizte is provided in Section III.A and
the specific augmentations in the Budget Justification).

A) Convert existing, proven data acquisition and reduction
nrograns from the PL/ACME larguage into Portran, consistent with
tine-critical assembly language programs for data acquisition and
instrument control. These programs will be written in Fortran to
enhance compatibility with the computer systems of other users of
such packages.

B) Modify these programs, 3S requirei, to handle acquisition and
reduztion of frequent or repetitive HRMS scans with selected
instrument performance feedback to the operator, and to take
2aivantage of the expanded capabilities of the extended 11/20
systam. Prototype GC/HRMS systems have heen developed at Stanford
and 2lsawhare, but this type of facility (in contrast to GC/LREMS)

YE
now available to the Stanford community. When this system
2lopad, service will be available to the Stanford community
searzth collaborators and, if our resources permit, to any
tist requesting assistance. In many instances this type of
llaboration will require far mora involvement of convergent
interests, efforts and skills than merely running samples on
request. We have in mind the chemical and eventually biological
interpretation of the analytical data as a matter of joint
concern, as appropriate.

2 WV
mo
yo

ahs bee
¥
mu
a

oO Ss

I
ie

7]
OQ

4

we have praviously illustrated the advantages of high resolution
Rass spectral data in the computer analysis of mass spectra {e.g.,
ref. 28). Also, we have previously shown that the same program
cai deal with analyses of mixtures without prior separation
especially when additional data (e.9., from selected metastable
Jlefocissiny experiments) were provided (Ref. 33). We wish to use
the MS rassurce and the comput2r program in further studies of
nixtures of compounds which are difficult or impossible to
separate by GC. The advent of routine systens for high pressure
Liquid chromatography have made many of these separations
possible, but the liquid chromatograph is not presently interfaced
to the MS.

Many of the problems of the user community require analysis of
complex mixtures which are amenable to treatment by GC/MS
techniques. We feel that where sample quantities permit,
acquisition of GC/HRMS data is highly desirable. These data can
be providai by the resource supplemented with computer support
{above).

We propose to continue tests of the GC/MS combination, operating
ander moderately high mass resolutions (5900-19000), to define in
Jetail the optimum operating conditions of the GC/HRMS
combination. This will provide the necessary information on
maximum practical sensitivity to be expected. This information
tan then he used in collahoration with the user community for
sampLo prenaration.

The 37/HRMS system would normally be operated at reduced mass
spactrometer resolutions to maximize sensitivity. We have
existing multiplet resolution programs to increase the resolving
power of the MS. We propose to provide the nultiplet resolution
program with heuristic quidance based on compositional variations
inferred from molecular ions or other singlet peaks. For example,
a cesolvyiny power of 10,000 is harely sufficient to resolve ions
which differ by CH2 vs. N (delta m = 0.012) for ions of about mass
193. Althsugh it will resolve CH4# vs. 0 doublets (delta m =
9.937) at this mass, it will not resolve closer doublets such as
T3N vs. H293 (delta m = 9.003). We can provide exhaustive
tabulations of multiplets hy mass separations (based on ref. 30)
which can be used by the muitiplet resolution program.

We have praviously indicated the power of metastable ion
information in the operation of our prograns for structure
alicidation {refs. 28, 33). Ye have extended one of our programs
(the MS predictor program) to propose metastable defocussing
experiments in order to avoid tolleaction of unnecessary data (see
Ain 2, below). Although we can collect these data (Barber-Elliott
techniyue) manually on our existing Varian MAT-711 MS, this is an

47
axcesedingqly wasteful operation, both in terms of sample
sonsumption and time. We propose to implement some automation of
collectian of these data on metastable ions. We also propose to
bazyin preliminary investigation of alternative modes of metastable
ion datermination {see Methods (Sec. ITI), helow).

2. Develop performance and theory formation programs to assist
in the solution of structure elucidation problems in biomedicine.

Tonputer programs have already been written for analysis of low
and high resolution mass spectra, for generation of acyclic and
cyclic molecular structures, for labelling structural skeletons
with atoms, for analyzing 13CMR spectra of amines and for
interpretation and summary of large volumes of data gathered on
molel tompounds {see Existing Capabilities above, for references).
We wish to increase the utility of these programs by providing
interactive facilities that allow easier access to them, by
increasing their generality ani power, and hy supplementing them
with new raasoning programs.

Performance Programs:

The current structure generator program will be subjected to
further datailed tests before using it for structure determination

problens.

A naw algorithm for generating cyclic skeletons (with
190 multiple bonds) will be projramned and checked. The algorithm
is written and informally proved. A formal proof will be devised
as wall. This algorithm represents one very powerful approach to
the problen of implementation of constraints, as discussed in the
following paragraph.

The generating programs will be modified to allow isomer
generation within constraints. Different kinds of constraints can
ba inferred from different kinds of spactroscopic data. We intend
to give tha program knowledge of a variety of these.

The Planner programs that infer constraints from mass
spectrometry data will he broadened to include additional
knowledge about the spectral hehavior of classes of compounds of
ralevance to the NIH-sponsored research of the user community. In
atdition, w2 will add the capability for utilization of
information ahont chemical isolation procedures (e.g., one expects
acidic and neutral compounds in solvent extraction of acidified
body fluids) and relative GC retention times (e.g., to admit the
possibility of homologous series).

We propose to implement a more general method for
inferring the idantity of the molecular ion whether or not this
appears explicitly in the spectrum. This information is important
for the successful operation of the structure generator and the
planiver. We want the program to use whatever information is
available and not depend, as it currently does, on having
knowledge of the structural class together with inference rules
foc that class.

Tnterface routines will be written to make it easier for other
scientists to use these programs. We have to wait for an

48
interactive system hefore starting this: AIM-SUMEX will he ideal.
Inpit/foutput routines will be crucial to easy use of the system.
Haywavec, we also want to give users the facility to understand the
system's reasoning steps so they can take advantage of it.

Tn addition to making the computer programs available through
ATM-SUMEX, we would like to translate parts of the LISP code into
another language - for reasens of both efficiency and
axportability. We have talked with computer professionals at IBM
Research Canter about using tha APL language. FORTRAN, ALGOL and
PL/1 are other languages whose merits for our purposes we will
explore.

Wea wish to continue a low-level of effort on computer
programs that interpret other kinds of spectroscopic data.
Planning programs Similar to the “4S Planner could be written for
automatic analysis of data fron other spectroscopic
tezhnigquas{e.q., IP, UV), as we have illustrated for 13CMR (ref.
39),

Tha structure generator's view of chemical struct ire is
topological and is presently unconstrained hy bond lengths and
anjl2s. Because stereochemical considerations are frequently
important in structure elucidation, we propose to begin
consideration of stereochemistry in the structure generation and
evaluation processes.

A proyram with detailed knowledge about information
abhtainable from various spectroscopic techniques could be written
to exanine a list of candidate solutions and propose experiments
necessary and sufficient to distinquish among them. The program
would represert an extended Predictor (e.g., ref. 27). We have a
ficst version of a program that suggests "crucial" metastable
peaks to he sought in order to distinguish among candidate
Structures. Work on this proqram will continue at a low level of
activity, possibly expanding into areas other than MS. One topic
w2 will continue to pursue is our collaborative effort with Dr.
Silda Loaw, Genetics Department, on th2 potential application of
molecular orbital theory to pr2adiction of mass spectra (ref. 71).

Theory Formation Proqrams:

The rile formation progran ({RULEGFEN) will be extended so that
i+ can search a larger space of rules. Present a priori
zoristraints on the rule generation give us a search reduction from
teas of millions to a thousand possible rules. Even though search
hairistics now allow efficient search o9f these possibilities, we
want to be able to deal with much larger spaces efficiently, as
whoa the number of primitive predicates is drastically increased.

The RULFGEN progran will be modified so that complex
Fragmentation and rearrangement processes are manipulated nearly
as easily as simple fragmentations. The program currently finds
frajmentation rules involving one or two bonds, possibly followed
by hydrogea migration. In the case of cyclic systems such as
astroaqens, however, the program must be able to work with sets of
threa or more bonds in some cleavages.

Interactive programs will he provided on AIM-SUMEX for the
investigator to guery the rule generation program. For example,

49
many questions now arise about the program steps by which the
program infers the rules it suggests as explanations of the
regularities. Why, for exanple, was some particular rule not
tonsideread plausible?

New data will have to be selected in order to test the rules
and to differentiate among competing rules. Wa will write a
program that suygests new experiments (i.e., new data to obtain),
depending on the nature of the existing rules.

The t2st phase of the theory formation program will be
written as an evaluation function of each rule against new data.
Tasofar as any new experiments are "crucial" experiments, the
avaluation function may merely reject a proposed rule. Mostly,
yoweaver, riles will have to be evaluated against new data along
many dimensions: frequency, strength of evidence, uniqueness,
simplicity, and the like.

ge wish to experiment with the whole theory formation program
to determine the critical aspects of our design. For example, {1)
how sensitive is the program t23 discrepancies, inconsistencies and
errors in the data? (2) how well can the program find rules
within a slightly different moilel of chemistry? (3) how well can
the program perform with one pass through the data, or several
passes? and (4) how critical are the principles of theory
Formation?

3. Apply the structure elucidation techniques - both
nstcumentation and computer programs - to biomedically relevant

in
compounds.

Jur own interests are in elucilating the structures of, and
anierstanding the MS of, marine sterols, hormonal steroids, and
compounds isolated from human body fluids that can be associated
with aqanatic disorders (from research in the GRC). In addition,
we will be working closely with members of the Stanford Medical
School and Chemistry Department - in particular those mentioned
ahove {Section I.B.4) - on their structure elucidation problems in
which MS will be used. Although most users expect to require HRMS
and SC/HRMS data, some of their problems will be attacked
ntilizing SC/LRMS techniques and library search through (usually)
restrict2d libraries of mass spectral data. We propose to
investiaate soma extensions to the technique of library search
{s2e Methois) to complement our existing and planned DENDRAL
preqrams. We plan to continue our exchange of mass spectral data
ana library search information as we have previously done with Dr.
S$. Yarkey (University of Colorado Madical School) and Dr. F. W.
McLafferty (Cornell University).

Rs in the past, attention to new biomedical research problems will
lead to increased capabilities in the computer programs. We
require close communication with the paople engaged in the
research so that the programs actually assist the researcher while
increasing in power. Collaborative proposals have come out of
suzh past DENDRAL sponsored work, for example, large portions of
the 3RC proposal and a proposal for 13CMR research.

W2 anvision the interaction and collaboration with the user
Sommunity to involve the following:

56