Technical Report No. IRL 1073

MOLECULAR BIOLOGY APPLICATIONS OF
MASS SPECTROMETRY

Final Report Covering Period July 1, 1965 to December 31, 1967
For
Air Force Office of Scientific Research
Contract AF 49(638)-1599

 

Instrumentation Research Laboratory, Department of Genetics
Stanford University School of Medicine
Palo Alto, California
MOLECULAR BIOLOGY APPLICATIONS OF MASS SPECTROMETRY

Final Report Covering the Period July 1, 1965 to December 31, 1967

Air Force Office of Scientific Research

Contract AF 49(638)~1599

Instrumentation Research Laboratory, Department of Genetics

2

Stanford University School of Medicine

Palo Alto, California

 

Joshua Lederberg i)

Principal Investigator
”)

Le

/ /
Elliott C. Levinthal \
Program Director

 
INTRODUCTION

The research objectives and purposes of our Air Force Program
entitled "Molecular Biology Applications of Mass Spectrometry" have been
closely related to those of our NASA program entitled "Cytochemical
Studies of Planetary Microorganisms - Explorations in Exobiology". This
is not too surprising since the same reasons that make mass spectrometry
a powerful tool for biological explorations carried out remotely on a
planet forty million miles distant from the earth make it a most sensitive
and selective method for analyzing organic molecules important in problems
of molecular biology germane to modern medicine. It is a potentially
powerful method for investigating the unknown structure of unknown mole-
cules innerve tissue that form the engram which is part of the process
of memory as well as Martian surface material. Cytochemistry via mass
spectrometry is still a distant and challenging goal. However, signifi-
cant progress has been made during this period. In addition to building
a base for further advances our efforts have yielded results of present
value.

The problem and the program can be subdivided into separate areas of
concern. First, there is the question of volatilizing the molecules of
interest. This can be approached by means of chemical modification of
the class of molecules under investigation. We have successfully applied
this concept to problems of resolution and identification of optical
isomers of amino acids using the combination of a gas chromatograph and

mass spectrometer. A second but much more difficult method, which has
the advantage that it is more directly applicable to the goal of
cytochemistry, could conceivably utilize electron, heavy particle or
photon beam energy. We have investigated the use of both heavy parti-
cles and laser photon beams. In the case of the former, we had useful
but discouraging results. In the latter case our present results show
some real promise.

A second subdivision of this effort addresses itself to acquiring
basic data on the mass spectra of a large number of monomers of biologi-
cal interest. A report covering the work on amino acids has already
been published. Work on nucleotides and related products is in progress.
Some of these results are referred to and discussed in this report.

Computer control of mass spectrometers describes the third cate-
gorization of the program. Full advantage of a mass spectrometer as a
biological tool can only be achieved when the instrument is under
computer control. The typical processes of calibration and optimization
of operating parameters are sufficiently complex that they require
automation if it is desired to analyze a large number of spectra in a
short period of time. We have published a report which describes a
very effective system for computer operation of both a time-of-flight
and quadrupole mass spectrometer. This system has been and is
continuing to be used for biological research purposes. The system is
being elaborated to provide a more sophisticated level of control. In
addition, work is underway to achieve some degree of control of high

resolution mass spectrometers that use magnetic filters.
Fourthly, there is the question of data retrieval for subsequent
computer analysis. The ultimate goal of a two dimensional micro-descrip-
tion of the distribution of molecules in a tissue by means of their mass
spectra presents formidable problems of data handling. The bandwidth
requirements are at least an order of magnitude greater than color video.
High bandwidth data retrieval and buffer storage are required. We have
not directly confronted this problem. General advances in the technology
of high speed solid state switching devices and information storage
methods lend some hope for the future. We have, however, made some
modest steps. We have implemented an interface system for a direct data
link from a high resolution mass spectrometer to an IBM 360/50 computer.

The fifth and last subdivision of the program really represents
most clearly the ultimate goal for which the previously described
efforts provide the technological tools. Ultimately the spectra acquired
must be analyzed. This requires computer manipulation of chemical
hypotheses. This poses a problem in both artificial] intelligence and
organic chemistry. A great deal of progress has been made in this
direction. Most of this research is supported by the Advanced Research
Projects Agency of the Office of the Secretary of Defense, Contract No.
SD-183, and carried out in collaboration with Professor E. Feigenbaum of
the Department of Computer Science.

This final report is divided into five sections appropriate to each
of the areas previously described. In most cases it will rely heavily

on references to papers previously published in professional journals
and technical reports which in some cases have been submitted to other
agencies. For the sake of completeness, some items are included in the

bibliographies which predated the Air Force contract.
I. Volatilization of Molecules of Biological Interest
a. Chemical Methods

Molecules of biological interest are characterized by asymmetries
at one or more of the carbon atoms incorporated in the molecule. From
the viewpoint of the exobiologist this statement is the basis of the
well-known significance of optical activity as a clue for the recogni-
tion of life. The preparation of volatile diastereoisomers, their sepa-
ration by gas chromatograph and their further identification by a mass
Spectrometer provides a method important to both terrestrial and
extraterrestrial biology. The general concept is elucidated in the
following papers enumerated in the bibliography which is part of this
section (references 1.1, 1.2). Initially the technique was applied
to the high sensitivity scanning of amino acids for optical activity
(references 1.3 to 1.8). Other papers demonstrate the general
applicability of the method (references 1.9 to 1.14).
b. Physical Methods of Volatilization

The goal of this aspect of the investigation was to develop a

method which ultimately will lead to mass spectrometry image scanning
and thus allow the possibility of microidentification and the physical
analysis of the cellular and subcellular constituents of biological
structures,

The following material is presented in more detail than other
sections of the report representing comparable efforts since it does
not yet appear in the appropriate professional journals. It has

however been described in the semi-annual status reports submitted
under our National Aeronautics and Space Administration grant No. NsG 81.
The initial efforts were based on the developments of Professor R.
Castaing and George Slodzian of the University of Paris who had
developed an imaging, secondary-ion-emission mass spectrometer which
produced magnified images of the surface distribution of atomic species
with one micron resolution. The intention of our efforts was to
ascertain if sufficient numbers of intact molecular ions characteristic
of organic particle materials could be evolved to enable the determina-
tion of the molecular distribution and structures of biological interest.
For these purposes Dr. George Slodzian joined our staff for a period of
approximately a year and a replica of the prototype developed by Castaing
and Slodzian consisting of an inert gas ionizer, primary ion gun,

target support assembly and secondary ion electrostatic lens structure
was acquired. The mass filtering was accomplished by use of a 60°

signal focusing mass spectrometer that was designed and constructed in
our laboratory.

Positive primary ions of argon and hydrogen were employed to
bombard polyethylene, nylon, polyvinylpyrolidone methylcellulose,
graphite, OFHC copper, gold, and aluminum. Both positive and negative
secondary ions were examined. The incident primary ion energy was
approximately 6 kev and 14 kev during the examination of positive and
negative secondaries, respectively.

The molecules evolved from the bombarded organic samples were
found generally to constitute rearranged configurations not representa-

tive of the original sample structure. It seems probable that the
incident primaries produce sufficient local dissipation of energy to
devastate the molecular structure. Consequently, the "hottest" of the
evolved secondaries come off as rearranged molecular ions and the
majority of the material recondenses on the surface, subsequently to be
re-evolved, retaining little information regarding the original structure.
As often happens there has been an unexpected finding in this work.
At one state in the data analysis it was decided to plot only those
mass peaks in any given run that corresponded to singly ionized integral
numbers of carbon atoms clumped together. It was thought that this
information might provide a useful clue as to the nature of the processes
taking place at the surface. When such plots were made for negative
secondaries, a distinct pattern revealed itself for all observed
combinations of primary ions and organic targets. Namely, there was a
preference for the negative secondaries to contain an even number of
carbon atoms. This effect has been seen to exist out to 12 carbon
atoms, whereupon the effect faded out into the "noise". The effect has
been found to persist with the addition of one or two protons to the
complex. When graphite was examined it was found that the curve
corresponding to one proton affixed to the complex was a replica of the
unprotonated curve but displaced downward, on a semilog plot. The
addition of two protons, however, so enhanced the effect that it could
be observed out to 14 carbons and over 6 decades of intensity where-
upon instrumental sensitivity limited further observation. A typical

odd-even enhancement in intensity was 10-fold.
A complementary but less pronounced enhancement of odd carbon
clumps over evens has been found for positive secondaries.

Arguments based on observed odd-even carbon effects have even
recently been offered in the literature in support of the biogenic
origin of oil; the rationale being that such preferential effects
mediate in favor of an ordered production mechanism. That odd-even
carbon effect can be generated by abiogenic mechanisms would appear
to weaken this particular argument. Odd-even carbon effects have been
reported in the past. They appear, however, not to be widely known.
Since the only explanation we can find for these effects calls for the
carbon complexes to be in the form of linear chains, we are forced to
conclude that most of the evolved ionized carbon complexes are in such
form.

Subsequently our efforts were directed to the investigation of the
suitability of laser induced vaporization as a mechanism for enabling
spacial resolution in mass spectro analysis of organic solids. The
additional experiments were conducted with an Optics Technology
Model 100 pulse ruby laser coupled to a Bendix time-of-flight (TOF)
mass spectrometer (MS).

Our measurements have indicated a lasing threshold of 190 joules
input with a laser output of 3.5 x 1072 joules at the maximum available
input of 440 joules. The wavelength of the laser radiation is 6943 A°.
Beam divergence at threshold appears to be in approximate accord with

the specified figure. We have some evidence that the full angle beam
divergence at maximum output is approximately 5 x 10°? rad; we have not
however, performed accurate determinations of the directional dependence
of the beam brightness of the radiation issuing from the laser.

Figure 1.1 illustrates schematically the optical configuration
utilized for these mass spectral studies. A portion of the radiation
issuing horizontally from the pulsed ruby laser 1 is reflected downward
by the beam splitter 2 toward the simple biconvex converging lens 3 of
focal length 100 mm, transmitted through the planar glass window 4 to
focus at the target 5 supported by probe 6 inside the source chamber in
the vacuum environment of the MS.

Alignment and aiming are accomplished by arranging that the quasi-
CW radiation from a 1 milliwatt Optics Technology Model 170 He-Ne gas
laser 7, reflected off mirror 8, partially transmitted through splitter
2, and reflected off the ruby and mutually parallel mirrors in the
laser head 1 returns to the point of origin at the gas laser 7. Since
this alignment ensures normal reflection of CW radiation off the laser
optics, it follows that the pulsed ruby radiation will be incident at
the same point on the target as that illuminated by the portion of the
CW radiation following the path 1-2-3-4-5. The point of impact of the

laser radiation is adjusted by horizontal translation of the converging
lens 3.

Each 100 usec the MS fires a 0.25 usec duration burst of ionizing
electrons transversely across the source chamber immediately above that

target. That portion of the material that is vaporized by the laser and

10
TT

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

©

FIGURE 1.1

Schematic of Laser Optical Configuration

Used for Preliminary Studies

  

PULSED RUBY
LASER

 

 

 

 

 
is positively fonized by each electron burst is impelled, by application
of a 2700 volt kick, into a linear field-free drift space at the

terminal end of which is a high speed ion detector.
The voltage kick V imparts to each ion a velocity v given

approximately by

 

Vv =V 2neV/m > (1.a)
where n is the number of electrons removed from the molecule, e is

the magnitude of the electronic charge, and m is the mass of the ion.
The elapsed time between application of the kick and detection of
arrival of an ion is given approximately by

t = L/v , (1.b)
where t is the transit time down the drift tube of length 2.
Substitution of (l.a) into (1.b) yields

t = RV m/ (2neV) . (1.c)

Conversion of the Mass-to-charge dependent velocity dispersion into ion

ke

time-of-arrival dispersion at the detector enables masses to be
identified from the cathode ray oscilloscope (CRO) display of ion
current versus time. The capture of maximum spectral information

over one or a succession of MS repetition cycles (several of the 100 p-
sec intervals) has involved photographing the CRO display of output.
Utilization of a procedure based upon the approximate quadratic
relationship between m and t expressed in (1.c) has enabled us to

associate mass numbers with recorded peaks out to the neighborhood of

400 atomic mass units (amu).

12
Figure 1.2 is a schematic representation of the initial version
of the pulsing circuitry utilized external to the MS to enable the
recording of either one or a series of successive spectra. The system
allows for the establishment of a controllable delay between firing of
the laser and the initiation of the CRO display. The number of
successive spectra displayed can be freely varied. Successive spectra
are vertically displaced from one another on the CRO. The joint
resolution of the CRO and the 10,000 ASA Polaroid film is insufficient
to enable unambiguous mass identification over the range 0-400 on a
Single trace. Recording of this mass range has generally required 5 to
10 laser shots, the mass range window for each shot being successively
displaced to higher masses.

The mass spectral analysis of the vapor produced by laser
irradiation of a powdered crystalline target of N-dinitrophenyl
(DNP)-L-isoleucine is presented in Table 1.1. There is also presented,
for comparative purposes in this table, the spectrum obtained by
conventional crucible warming of the same sample to 40°C. The structural
and graphic formulae for this sample are presented in Figure 1.3.

The photographic recording process and subsequent identification
and evaluation of mass peaks is at present a time consuming process
that does not readily lend itself to rapid scanning of a heterogeneous
target. This limitation is for the moment, however, academic. We
found that this initial experimental laser configuration appeared to

deliver insufficient focused pulsed areal energy density to vaporize

13
®

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LASER
BENDIX MANUAL
M NN NAN
TOF MS ent TRIGGER
SCOPE
TRIGGER
OUTPUT
Co) y THT RAMP GEN Q w |
Oy 1342 (TEK 162) (a) |
O _™
LEVEL Sens. ] ()
ADDER | © | PEDESTAL GEN} —»(¥! [MONITOR CRO
_I LL] (TEK 163) 7 (TEK 360)
(3 42 ,

 

 

O} aft

 

 

 

 

 

 

 

 

 

er

 

 

13 42

13 42

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(z)

RAMP GEN ©)
: 3 3 I Nf

TRIGGER
= --——— oo
|
| l
| |
| IA 4 13Né@/2 |
| KA |
©) ADDER Y
| CRO VERT. | (TEK 1A |
| | (TeK 450 (TEK IAN) |
a |

MS OUTPUT befDehn,
a
FIGURE 1.2

Block Diagram of Pulse Circuitry Utilized to Enable Recording

14

of Either One or a Series of Successive Mass Spectra

 

 
ST

Table 1.1. Relative Intensities of Laser Spectrum and Crucible (40 Deg. C) Spectrum
of dinitrophenyl-L-Isoleucine.

Mass Laser Cruc. Mass Laser Cruc. Mass Laser Cruc, Mass Laser Cruc. Mass Laser Cruc. Mass Laser Cruc.

1 0 1 51 14 1 101 10 4 151 0 0 201 0 16 251 0 0

2 0 1 52 25 1 102 rk 1 152 0 0 202 20 3 252 72 5

3 0 0 53 23 1 103 27 1 153 0 0 203 0 0 253 0 0

4 0 0 5h 22 0 104 17 0 154 0 0 206 0 0 254 0 0

5 0 0 55 40 2 105 7 0 155 0 0 205 0 0 255 0 0

6 0 0 56 8 0 106 5 0 156 0 0 206 24 0 256 0 0

7 0 0 57 100 3 107 8 0 157 0 0 207 0 0 257 0 0

8 0 0 58 7 0 108 0 0 158 0 0 208 0 0 258 0 0

9 0 0 59 11 0 109 0 0 159 0 0 209 0 0 259 0 0
10 0 9 60 0 0 110 0 0 160 0 0 210 0 0 260 0 0
11 0 0 61 3 0 111 0 0 161 0 0 211 0 0 261 0 0
12 0 0 62 10 0 112 0 0 162 0 0 212 0 0 262 0 0
13 0 0 63 38 1 113 0 0 163 0 0 213 0 0 263 0 0
14 3 5 64 28 0 114 8 Q 164 10 0 214 0 0 264 0 0
15 4 0 65 1k 0 115 0 0 165 0 0 215 0 0 265 0 0
16 0 5 66 8 0 116 0 0 166 ll 1 216 i) 0 266 0 0
17 3 20 67 14 0 117 12 0 167 0 0 217 0 0 267 0 0
18 14 100 68 8 0 118 3 0 168 0 0 218 0 0 268 0 0
19 0 0 69 4S 1 119 12 0 169 0 0 219 0 0 269 0 0
20 0 0 70 7 ] 120 2 0 170 0 0 220 0 0 270 0 0
21 0 0 71 i] 0 121 0 0 171 0 0 221 0 0 271 0 0
22 0 0 72 0 0 122 5 0 172 0 0 222 0 0 272 0 0
23 0 0 73 5 0 123 0 0 173 0 0 223 0 0 273 0 0
24 0 0 7h 12 0 124 0 0 174 0 0 224 13 0 274 0 0
25 0 0 75 48 2 125 0 0 175 0 0 225 0 0 275 0 0
26 0 0 76 19 1 126 0 0 176 0 0 226 0 0 276 0 0
27 46 2 77 20 1 127 0 0 177 0 0 227 0 0 277 0 0
28 67 100 78 20, #1 128 0 0 178 0 0 228 0 0 278 0 0
29 +100 7 79 0 0 129 0 0 179 0 0 229 0 0 279 0 0
30 62 1 80 0 0 130 7 0 180 0 0 230 0 0 280 0 0
31 12 1 81 0 0 131 7 0 181 0 0 231 0 0 281 0 0
32 28 37 82 11 0 132 3 0 182 0 0 232 0 0 282 0 0
33 0 0 83 0 0 133 0 1 183 0 0 233 0 0 283 0 0
34 0 0 gh 7 0 134 14 1 184 0 1 234 0 0 284 0 0
35 0 0 85 0 Q 135 0 0 185 0 0 235 0 0 285 0 0
36 0 0 86 0 0 136 0 0 186 0 0 236 0 0 286 0 0
37 0 0 87 5 0 137 0 0 187 0 0 237 0 0 287 0 0
38 0 0 88 0 0 138 0 0 188 0 0 238 0 0 288 0 0
39 55 2 89 0 0 139 0 0 189 0 0 239 0 0 289 0 0
40 22 2 90 11 0 140 0 0 190 0 0 240 8 1 290 0 0
41 100 7 91 0 0 161 0 0 191 0 0 2h1 0 0 291 0 0
42 18 1 92 0 0 142 0 0 192 0 0 242 0 0 292 0 0
43 24 2 93 0 0 143 0 0 193 0 0 243 0 0 293 0 0
4h 0 1 9h 0 0 164 0 0 194 0 0 24h 0 0 294 0 0
45 37 1 95 0 0 165 0 0 195 0 0 265 0 0 295 0 0
46 0 0 96 0 0 146 0 0 196 22 1 246 0 0 296 0 0
47 0 0 97 0 0 147 7 0 197 0 0 247 0 0 297 0 1
48 0 0 98 0 0 148 7 0 198 0 5 248 0 0 298 0 0
49 0 0 99 10 3 149 0 0 199 0 8 249 0 0 299 0 0
50 22 0 100 0 3 150 0 0 200 0 12 250 0 0 300 0 0
EMPIRICAL FORMULA OF (DNP)-L~- ISOLEUCINE: Cio His Og Nz

H\— ¢ —<“CH3 _—
NI7 = mY
26 — ——_—_ 4 — 720 45
240 H-N—C —€ =
N \
| ( HO
| H Nw -
—~—— — 252
—_— oo
/ H yer NOo >
NY a
130 C c \
/\67 | II
\ He /o™H /
c
\ hop
\ 2 “
~~ __ ee

GRAPHIC FORMULA OF (DNP)-L- ISOLEUCINE

FIGURE 1.3

Empirical and Graphic Formulae for (DNP)-L-Isoleucine.
The dashed lines suggest non-rearrangement subgroupings of

atoms that might give some of observed mass numbers.

16
many potentially interesting materials. We then undertook a re-
instrumentation designed to rectify this deficiency.

Our reinstrumentation was intended to accomplish three purposes:

(1) Reduction of the focused spot size;

(2) More precise aiming of the laser;

(3) An increase of the areal energy density delivered
at the focus of the laser beam.

The new system, illustrated in Figure 1.4, has been constructed
and is now in use. The design optimizes delivered energy density, with
full energy delivery, by maximizing the solid angle of condensed
radiation. Practical constraints necessitate a sufficiently long
distance from the condensing lens to the working point that a lens
aperture larger than the laser beam is required. The laser beam is
expanded to fill the condensing lens by placing a diverging lens
immediately in front of the pulsed ruby laser.

The system delivers approximately a 1 millisec ~ 10 millijoule
pulse of ruby radiation to an approximately 70 micron diameter spot.
The present optical configuration involves the placement of a planar
glass vacuum port within the cone of rays converging to the target. A
significant contribution to the observed spot size derives from the
sperical aberration contributed by the window.

We have found that the optical pulse is of insufficient magnitude
to vaporize bulk reflective samples. Such samples are handled by

depositing them in such a manner on a black oxidized copper support

17
 

 

— _
=_—oe—

 

 

 

PULSED RUBY
LASER

@ ©

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ore oe

 

 

 

 

 

)

 

 

 

 

 

 

 

 

 

 

 

 

FIGURE 1.4

Schematic of Laser Optical System

18

 
that thermal vaporization is accomplished by thermal transfer from the
copper to the sample.

Placement of the sample at the point of concentration of the laser
radiation is achieved with mirror 7, in Figure 1.4, rotated to its
phantom position. The sample target point is placed so that it is
imaged at the cross hair 9 of the eyepiece. The optical configuration
is such that the portion of the sample imaged on the crosshair will be
at the focal point of the condensed laser radiation when the system is
fired. |

Figure 1.5 illustrates the probe that has been constructed to
provide the three degrees of freedom required to properly position the
sample. The probe consists of a hollow tube containing a centrally
located stem penetrating through a 0.010 inch thick stainless steel
diaphragm brazed to the stem and the tube. The sample is placed on the
upper end of the removable stem tip. The shoulder on the probe base
structure seats on the vacuum lock of the mass spectrometer source. The
sample is raised or lowered relative to the shoulder by rotating a
concentric wheel engaged to the threaded base of the tube through an
intermediate annular key. The sample is displaced perpendicular to the
tube axis by two mutually orthogonal micrometers that pivot the stem
about the flexible diaphragm. The micrometer driven motion is reduced
10:1 at the sample. The diaphragm tolerates 2 x 10-2 radians of

angular displacement from equilibrium before acquiring a set.

19
 

 

EA
H
MV] 4]
Vj a
|
, M
1
ly
Vj
4
h
U
4
VY]
y
Vy
yj
Ly ———-_-_—-—_—_1
\ , ONE INCH
Vy iY)
y
Vy
Vy
i

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

FIGURE 1.5
Mass spectrometer sample probe

for laser induced vaporization

20
Figure 1.6 is a schematic illustration of a control unit now being
constructed to expedite the recording of output from the laser-mass
spectrometer system. When the start button is pressed with the unit
in “normal” function, a "background" number of reference gas traces is
recorded across the top and then again across the bottom of the scope
face, the laser is then fired, and after the designated number of mass
spectrometer "cycles delayed" (100 microsec/cycle), the "sample" number
of successive cycles of mass spectral analysis of the laser induced
output is recorded in an equally spaced display between the upper and

lower background traces.

21
 

The top and bottom traces show the background,
The middte traces are two successive mass spectra
of the sample which has been volatized by the LASER.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ext. Trig.
O cx. A O
©) BENDIX TOF he \ LASER >
cH. B
aR” O-
nh
7 RECURRENT CYCLES
BACKGROUND DELAYED SAMPLE
—— SINGLE START
~~ NORMAL C 3 e O
N LASER ONLY
TRIGGER BASE
BENDIX DELAY LEVEL
MASTER PULSE SCOPE FIRE LASER
TRIGGER CH. B
SO O Oo:

 

 

 

 

 

 

 

 

 

FIGURE 1.6

Control unit to expedite scope display of laser induced mass spectra

22
REFERENCES

1.1. B. Halpern, J. W. Westley, E. C. Levinthal and J. Lederberg,
"The Pasteur Probe: An Assay for Molecular Asymmetry: Life
Science and Space Research, North-Holland, Amsterdam (1967).

1.2. J. W. Westley, "Detection of Optical Activity as a Sign of Life."
presented to the Amer. Astro. Soc., (1966). IRL-1047.

1.3. 8B. Halpern and J. W. Westley, "High Sensitivity Optical Resolution o
of Poly-functional Amino Acids by Gas Liquid Chromatography."
Tetrahedron Letters 2283-6 (1966). IRL 1043.

1.4. 3B. Halpern, P. J. Anderson, J. W. Westley and J. Lederberg, "Demon-
stration of the Stereospecific Action of Microorganisms in Soil
by G.L.C."" Analytical Biochemistry (1966). IRL 1050.

 

1.5. B. Halpern and J. W. Westley, "High Sensitivity Optical Resolution
of D,L Amino Acids by Gas Chromatography.'' Chem. Comm. 12: 246
(1965).

1.6. B. Halpern and J. W. Westley, "High Sensitivity Optical Resolution
of D,L Amino Acids by Gas Chromatography.'"' Biochem. Biophys. Res.
Comm. 19(3) 361 (1965).

 

1.7. B. Halpern and J. W. Westley, "Resolution of Neutral D,L Amino
Acids via Their L-Menthyl Ester Derivatives." Chem. Comm.
18:247 (1965).

a
1.8. B. Halpern, J. W. Westley, I. vonWredenhagen and J. Lederberg,
"Optical Resolution of D,L Amino Acids by Gas Liquid Chromatography
and Mass Spectrometry." Biochem. Biophys. Res. Comm. 20:710 (1965).

1.9. B. Halpern, J. Ricks, and J. W. Westley, "The Stereospecificity of
«-Chymotrypsin-catalysed Reactions." IRL 1042.

1.10. B. Halpern, J. Ricks, and J. W. Westley, "The «-Chymotrypsin-
catalysed Hydrolysis and Alcoholysis of Specific Ester Substrates
in the Presence of Added Nucleophiles." IRL 1045.

1.11. B. Halpern and J. W. Westley, "High Sensitivity Optical Resolution

of Amines by Gas Chromatography." Chem. Comm. 2:34 (1966).
IRL 1041.

23
1.12.

1.13.

1.14.

B. Halpern, J. Ricks, J. W. Westley, "The Stereospecificity of
a-Chymotrypsin-Catalyzed Hydrolysis and Alcoholysis of Specific
Ester Substrates." Aust. J. Chem., 20:389 (1967). IRL 1049.

B. L. Karger, R. L. Stern, W. Keane, B. Halpern, and J. W. Westley,
GLC Separation of Diastereoisomeric Amides of Racemic Cyclic
Amines." J. Anal. Chem. 39:228 (1967).

B. Halpern and J. W. Westley, "Chemical Resolution of Secondary

(+) Alcohols." Aust. J. Chem. 19(8):1533-1534 (1966).
IRL 1044.

24
II. Spectral Analysis of Monomers

In a preliminary study to provide insight into operational
problems of computer control a comprehensive set of spectra of amino
acids, as observed with solid samples in the Bendix time-of-flight
instrument, was collected. This work is described in a technical
report prepared by Miss Nancy Martin?*}, This work has been supported
jointly by a National Aeronautics and Space Administration Grant NsG 81-
60; National Institutes of Neurological Diseases and Blindness, Grant
No. NB 04270; and Air Force Grant No. AF~AFOSA-886-65. This Air
Force Grant was the predecessor to the present Air Force Contract for
which this final report is being prepared.

One of the most important problems of data acquisition, the cali-
bration of mass numbers, had only begun to be handled by the computer
system at this stage of development, and assignments given in this
report must be regarded as tentative. Furthermore, no attempt was
made to assess the ultimate sensitivity of the assay. Nevertheless,
these data showed the potentialities of the technique, especially when
the distinctive temperature characteristics of each amino acid are
considered.

We commenced the mass analysis of selected organic samples with
the completed laser optical system described in the previous section.
Our first results demonstrated that we were readily able to detect the
announced apparent molecular peaks for the four nitrogenous bases -
adenine, guanine, thymine, and cytosine. We also demonstrated that the

base peaks are conspicuous for the nucleosides deoxyadenosine,

25
deoxyguanosine, and deoxycytidine. We also found peaks at the base
masses in the two nucleotides we have so far examined, deoxyguanosine.
monophosphate and thymidine monophosphate.

Table 2.1 is a listing of the spectra that have been obtained and
are available in the computer data file for display and comparative
analysis. Section IV of this report includes a discussion of this

aspect of the work.

REFERENCES

2.1. N. Martin, "An Investigation of the Mass Spectra of Twenty-Two
Free Amino Acids." IRL-1035, September 21, 1965.

26
Table 2.1

B-Subtius - DNA

Salmon Sperm DNA - Crucible
Adenylic Acid - Bendix Crucible
Deoxycytidine - Bendix Crucible
3M-FC43 - Bendix Crucible
Deoxyguanosine ~ Bendix Crucible
Deoxyadeninosine - Bendix Crucible
Deoxyadenosine Monophosphate - Crucible
Thymidine Phosphate - Crucible
Thymine - Crucible

Cytosine - Crucible

Guanine - Crucible

Adenine - Crucible

Deoxycytidine - LAS

Deoxyguanosine - LAS
Deoxyadenosine - LAS
Deoxyadenosine Monophosphate - LAS
Deoxyguanosine Monophosphate - LAS
Thymine - LAS

Thymidine Phosphate - LAS

Cytosine - LAS

Guanine ~ LAS

Adenine - LAS

Salmon Sperm DNA - LAS

P,0S - LAS

Dextrose LAS

Clean Copper Probe

27
III. Computer Control of Mass Spectrometers

The question of computer control of mass spectrometers is central
to our ultimate objective. The computers used in this part of the
program are a LINC and an IBM 360/50. The Instrumentation Research
Laboratory was assigned a LINC computer in 1963 as a participant in the
LINC evaluation program under the sponsorship of the NIH, NASA, and
U.S. Air Force. This was administered under an NIH grant FROO151-01.
The IBM computer is part of the Advanced Computer for Medical Research
(ACME) program supported by the NIH, Division of Research and Facilities
Resources under Grant FROO 311-01. We have applied these computational
and computer control facilities to a Bendix time-of-flight mass
spectrometer, an Electronic Associates, Inc., quadrupole mass spectrom-
eter, and an MS-9 high resolution mass spectrometer.

References 3.1 and 3.2 describe the development of some of the
software which has been used in this research.

The design and construction of hardware and software for an ACME
360-LINC communication gave us the capability of processing mass
spectrometer data gathered by the LINC on the 360/50. This overcame
the handicap posed by the limited computational ability of the LINC.
This limitation results from the LINC having a 12 bit integer arithmetic
system. Most data manipulation steps, therefore, must be done with
software routines for double-precision, floating-point arithmetic. The
result was that arithmetic operations took place at millisecond rather

than microsecond speeds.

28
With this facility we used the 360 to transfer the data collected
from the Bendix TOF mass spectrometer into a more usable form. A
computational process which required 20 minutes to complete on the LINC
was found to require only two minutes when done under the ACME time-
sharing system on the 360. In this case the data was sent to the 360,
operated on in allocated "time slices",and returned to the LINC for
display on the LINC oscilloscope. About half of the two minutes
required was the result of the tape reading and writing operations on
the LINC.

A computer interface and computer operating system for the
Electronic Associates Inc. (EAI) QUAD 300 mass spectrometer was
completed in June 1967. In this system the computer exercised control
and acquired spectra from the QUAD 300 quadrupole mass spectrometer.

It allowed the analysis of gas chromatograph effluents using a mass
spectrometer as a detector. The scientific results of these efforts
are discussed in the first section of this réport and the associated
references.

The details of this computer controlled gas chromatograph-mass
spectrometer system are separately reported (3.3). The instrumentation
consisted of a quadrupole mass spectrometer, a special small rack of
electronics, which were termed the "interface", a LINC computer, and a
computer software system developed in our laboratory. The following is
the abstract of that report.

"A mass spectrometer-computer system has been devised to

utilize the decision-making capabilities of the modern digital

29
computer. The system described assists the researcher

user by allowing a computer to query the researcher for

operating parameters. The computer translates these into

detailed control functions that operate the instrument,

The data acquired from the mass spectrometer is made

available to the researcher in an on-line system. The

system employs a small digital computer and an integer

resolution quadrupole mass spectrometer. A reference gas

is valved into the mass spectrometer by computer control

to permit automatic calibration. Spectra processing of

GLC effluent was demonstrated; the means and results are

given."

The design of the QUAD 300 computer interface made provision for
the use of the electronic hardware and the computer programs with the
Bendix time-of-flight mass spectrometer as well as the QUAD 300. In
addition to its use with the QUAD 300 this interface equipment was
installed with the Bendix TOF.

_A voltage boosting circuit was installed in the Bendix analog scan
unit to control the Bendix from the computer interface. This and other
rework on the Bendix was done in a manner to allow either conventional
or computer operation.

The Bendix time-of-flight, could be operated from and by the
computer in a manner virtually identical to that accomplished on the

QUAD 300. The system allows a reasonably fast data acquisition,

30
nominally 4 to 8 seconds per spectra, and an on-line plot or
presentation of the spectral measurements. The data acquisition time
is compatible with obtaining mass spectra during the period of sample
peaks in a gas chromatograph effluent. However, as yet no gas
chromatograph has been connected to the TOF mass spectrometer.

The completed system of computer control, data acquisition and data
presentation is very economical of the mass spectrometer operator's
time. It does allow more reliable data acquisition in a fraction,
perhaps as little as one tenth, of researcher and technician attention
to the mass spectrometer, chart interpretation, and manual data
reduction.

Some progress has been made on a project for direct computer
acquisition of high resolution mass spectra from a MS-9 mass spectrom-
eter and for eventual computer control of the system. This included
the installation of a remote data terminal (270-Y) of the ACME 360/50
system adjacent to the MS-9 mass spectrometer in the Chemistry
Department.

- The ACME IBM 360/50 is a time-sharing computer located in the
Medical Center and has a short cable connection to the Instrumentation
Research Laboratory. The MS-9 is physically located in another building
1500 feet (cable length) away. The 270X-Y is a special IBM remote data
access system designed to operate at these distances. The intent of
the MS-9 development is to take the electrical output of MS-9

detector and, via suitable conversions and transmissions, sample it in

31
real time with the ACME computer.

The required 270X~-Y system has just recently become usable from
the ACME terminals. It must still be tested at the rate of data
transmission required for the MS-9 operation. The ACME software system
must make provision for allocation of the large data storage space
required. Both of these requirements are within the specifications of
the respective systems. The first, the technical ability of the hardware
to achieve the speed, is not expected to cause any problems. The
software system necessary to accommodate this data within the framework
of time-shared computer use is not as straightforward. This software

system is still under development.
REFERENCES
3.1. R. K. Moore, “An Operating System for the LINC Computer".

Technical Report No. IRL-1038, 1965.

3.2. R. B. Tucker, T. Coburn, W. Reynolds, and J. Bridges, "Software
for the LINC Computer". Technical Report No. IRL-1055, 1967.

3.3. W. E. Reynolds, J. C. Bridges, T. B. Coburn and R. B. Tucker,
"A Computer Operated Mass Spectrometer System". Technical
Report No. IRL-1062, 1967.

3.4. ACME Progress Report to NIH, October 1, 1966 - April 1, 1967,

NIH, Division of Research Facilities and Resources under Grant
FROO311-01.

32
IV. Data Retrieval and Display

Much of what we have done concerning the problem of data retrieval
is covered in the previous sections of this report and in the references
cited in those sections. The references cited in this section are for
background purposes only and predate the period of this contract.

We have not directly confronted the problems posed by the
requirement to handle a very high bandwidth data base that would be
generated by the ultimate mass spectral image scanning applications we
envisage. We have, however, designed, constructed and used systems which
allow us to explore with some efficiency small sections of this stream
of data. Although it is without the elegance that we ultimately seek
and requires a great deal more effort than might be desirable, this
does permit an evaluation of the suitability of other elements of the
system such as the method of volatilization and the mass spectrometer
itself for ultimate biological applications.

Mass spectra from the Bendix TOF are sent through a logarithmic
amplifier then digitized and stored on magnetic tape by the LINC
computer. A transformation is applied to the time axis of the spectra
to place the mass positions at equal intervals on this time axis.
Because of certain drifts in the TOF, the mass values of this altered
spectrum cannot be accurately determined simply by their position on the
new scale. A method was developed for determining their position
which was quite time consuming with the limited computational abilities
of the LINC and will have to wait for the implementation of the ACME

system for practical usage.

33
A method has been developed for identifying integer mass peaks by
utilizing the display scope of the LINC computer to display a portion of
a spectrum and a raster of the peak positions. This method displays a
portion of the spectrum (after the rough linearization) on the screen
with a generated raster indicating the approximate separation of the
mass positions (see Figure 4.1). The raster can be stretched and
translated using the potentiometers on the LINC console to fit the peak
positions of the displayed spectrum. This approach combines the user's
ability to distinguish meaningful peaks with the computer's capacity to
calculate the raster, store the digitized spectrum, and store the
information about the mass peaks once the user has adjusted the

raster to fit the particular part of the spectrum being displayed.

Cece bacar dasa s bees ade ss ebeo nant eanstasus

 

FIGURE 4.1

34
The program is operated under the locally written LOSS monitor
system (see reference R. K. Moore,"An Operating System for the LINC
Computer", Technical Report No. IRL-1038) and uses its conventions for
handling the data tape (the spectrum). Upon loading the program, the
first portion of the spectrum (actually the transformed spectrum as
described above) is read into the memory and displayed on the oscillo-
scope (8 cm by 8 cm display area). About 40 mass positions are within
view of the user. Initially the user positions an illuminated "pointer"
over a known mass peak and enters the associated mass value using
potentiometers on the console. Lifting a console switch identifies
this position in the spectrum with the particular mass positions equal
to zero modulo five being enhanced. Thus a raster is developed for the
set of peaks based on the particular position and mass number entered
through the potentiometers. The distance between the elements of the
raster may now be expanded until the elements and the mass peaks
coincide. This expansion takes place about the reference mass previously
mentioned so as not to disturb its position. The user can, however,
translate the entire spectrum to improve the matching between the raster
and the observed mass peaks by adjusting another potentiometer.

With the matching accomplished the researcher can now either have
the amplitudes at the mass positions typed out (still in logarithm form)
or, by setting a switch recorded on magnetic tape for further processing
(see below). The mass positions considered in this operation are those

between a set of vertical bars on the screen which act as parentheses

35
around the masses under consideration. The user has complete flexi-
bility in determining the portion of the displayed peaks to type or
record or may investigate an area more than once to observe changes
due to altering the raster. In general, the latter will yield little
change since the program searches half a peak on either side of the
raster position to find the maximum in the signal.

Having investigated the displayed portion of the spectrum the
user moves the reference mass up to a position in the rightmost quarter
of the screen by moving the pointer to a mass position on the raster and
lifting a toggle switch. Lifting another switch will move the data on
the rightmost quarter of the screen to the extreme left and read in the
next portion of the spectrum. The raster can agadn be adjusted and
another set of mass values typed or recorded.

In actual practice it has been found that only a slight amount of
raster adjustment is needed (less than a peak width) in any one portion
of the spectrum and thus the user can move along quite rapidly. This
latter condition prompts one to think about automatic adjustment of the
scale using an algorithm which examines the spectrum to find the spacing
for each successive portion. While this would work well in the lower
part of the spectrum (say up to mass 75), the large voids in the higher
mass regions make these methods undependable. Once familiar with the
system, a user can cover the mass positions up to 300 in three or four

minutes if the output is being stored on tape rather than typed.

36
Once the data is on magnetic tape in the form of mass number and
amplitude number pairs, it is readily usable in a number of ways. It
can be put on standard seven channel digital tape or possibly sent
directly to another computer as intended under the ACME concept.
Presently the data is used as input to another LINC program which takes
the antilog of the mass amplitudes and produces a bar graph (see
Figure 4.2) in which the largest peak is considered to have value 100
and all other peaks are scaled accordingly.

We are commencing to use the Stanford Medical Center IBM-360 ACME
time-shared computer system for the acquisition, storage, retrieval,
manipulation, visual display, and comparison of mass spectral data.

Mass spectra obtained by laser induced vaporization of solid
samples are supplemented, when possible, with the spectra that result
from heating the samples in a quartz crucible. The photographic records
of the laser induced spectra are visually interpreted and manually
transcribed into the computer. The data obtained by crucible heating
is generated slowly enough that it may be readily tape recorded. In
the latter case a LINC computer is currently used to convert the tape
recorded information into a table of mass number vs. intensity, and this
information is then transmitted to ACME. We are about to start
operating, in the crucible mode, with a computer generated mass scan.

We have been interested in experimenting with the visual comparison
of spectra. We have just completed a program which enables this
comparison to be performed by observation of a television display driven
by ACME. Figure 4.3 shows the display format that appears when the

program is first called.
37
ge

yeeqg Jsesiey Jo quaoisg

100-

715

wa
Qo
}

25 +

 

 

 

 

 

 

| il th ‘|

 

bad tiilins sz
ARARALAI ALADAAALAS ALALLAL ELS ALB LEAL

iT i) qi iT thr iT. ir. shiz iT
PARRA LEA AS RAAAAL AL AT AAASLOA A) BALA LAED DEL EALAL SL | PY VEVENT TUT YY PET ETEPETESTPTS TPE ETLISTT OT YOSPOUTEVESTOTOTOTVYT TYPO FOUN YU OI POET TIE US CEL EL ES CTL YONETUUT EP UVT ENTLY PFO FUTYTYT IVIL TUTTE LITT ITY TT TITY

50 7 100 150

Mass to Charge Ratio

FIGURE 4.2
Mass Spectrum of Phenyl Alanine Methyl Ester HCL

200
 

 

|

\ 0-400 So

 

 

N f

 

 

'
4OO..

 

 

 

yoo.”

 

 

lw NS hee oe

 

 

 

 

 

DISFLAT \

A

 

 

 

 

 

 

 

 

 

 

@)

tS

 

FIGURE 4.3

39

 
The image consists of a square pattern divided up into a set of
rectangular zones. The straight line 16 is the common base line for two
spectra that may be simultaneously displayed. The mass peaks for the
spectrum displayed in region 4 will extend upward from the base line and
for the spectrum displayed in zone 5 they will extend downward. For each
of the two line spectra the mass numbers increase to the right. The numbers
in zones 6 and 8 denote the lower and upper limits, respectively, of the
mass range for the upper spectrum. The numbers in zones 7 and 9 serve a
similar function for the lower spectrum. The numbers in zones 2 and 3
identify the record numbers of the upper and lower Spectra, respectively.

All control in execution is accomplished by the use of a deck
mounted gearshift-like control with a sensing button on the end of it.
For historical reasons, this unit and the blinking spot 1 that it
positions on the display are referred to as "mouse". The horizontal and
vertical coordinates of the spot 1 on the screen are sensed by potentiom-
eters coupled to the manual control. Reading of these coordinates occurs
when the depressed sensing button is released.

When it is desired to select or change either a spectrum record
number or the mass limits, mouse is moved into the area consisting of
the set of zones 10, numbered 0 to 9. A multiple digit number may be
collected from region 10 by successive selections and mouse then used to
deposit this collected number in zone 11 for checking. The flashing of
bar 14 advises the user that the computer is waiting for mouse. This
information is useful when the computer is in heavy use. The selected

number may be either corrected or transferred to any one of zones 2, 3,

40
6, 7, 8, or 9 by directing mouse to the desired zone and releasing the
button. After selection is made, mouse may be used to sense zone 12,
labeled "DISPLAY", thereby signaling the computer to display the
selected spectra over the chosen mass ranges. The photograph reproduced
in Fig.4.4 demonstrates the comparison of the base guanine obtained by
use of the laser, record 31 displayed up, and the guanine spectrum
obtained by use of the crucible, record 35 shown down. Sensing mouse
above the base line and at a particular mass peak causes the "V'"'-like
symbol 17 to appear at the mass peak and the value of the mass peak, 151
in this case, to be displayed at location 18. Mass 151 is the guanine
molecular peak. Sensing below the base line causes the inverted "Vv"
denoted 19 to appear at the mass peak and the associated mass number 202
to appear at location 20. Mass 202 is one of the mercury isotopes
arising from the use of mercury as a pump fluid.

The photograph reproduced in Fig. 4.5 helps to illustrate some of
the options available to the user. Record 37, salmon sperm DNA
volatilized by the laser system, is displayed over the mass range 0-400 i
in the upper spectrum. The lower spectrum is a portion of the same
spectrum showing in further detail the masses ranging from 100 to 150.
The base peaks for cytosine (mass 111), thymine (mass 126) and adenine
(mass 135) are present. Guanine evidently got lost in the noise on
this shot.

Figure 4.6 illustrates the manner in which the display shown in

Fig. 4.5 was obtained. Spectrum 37 in the mass range 0-400 was first

41
 

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

=
151.
8],
Q. ¥O00.
0. 4yo0.
i
35, M
l
T
202. >
\
N
iE G1 1A 2] 30] Het Se 6. | N 8. | 9. *OISPLAY

 

 

 

 

 

 

 

 

 

 

 

 

 

FIGURE 4.4

42

 

 
 

 

0-400

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

43

37,
0. | | yoo,
100. | | | 150.
L
37, ‘
1
T
$
a)s.fa|sJudsielrlets DISPLAY
FIGURE 4.5

 

 
 

 

 

0-400

 

 

 

 

 

 

 

 

 

 

 

3 Ep
i TA ~A :

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

URE 4.6

44
displayed both up and down. Mouse was then sensed in region 13,
labeled "LIMITS". Following the sensing of LIMITS, mouse was sensed
in zone 5 at that mass position intended to be an upper limit for the
lower bound on the desired expanded mass range. An enhanced bar such as
the one shown at 21 then appeared at the position chosen by mouse.
Sensing mouse in the lower limit zone 7 then placed the newly selected
lower limit in this zone. Because of the discreteness of available
point sites on the digitally driven television screen the lower limit is
automatically rounded to the nearest multiple of 25 that does not exceed
the mass identified by the enhanced bar. Mass 100 resulted. An "L"~
like symbol, denoted 22, is created to show the mass location of the
new lower bound on the future spectrum. A flashing "X" appears at 24
to tell the user that the displayed spectrum is no longer valid.
Sensing mouse successively in the LIMITS zone, in the mass zone, and in
the high limit zone provides the new upper limit 150.

Sensing DISPLAY then results in the appearance of display shown in
Fig. 4.5. If mouse is subsequently sensed in zone 13, labeled "0-400"
the upper and lower limits of either of the spectra may be restored to
the range 0-400 by a subsequent sensing of mouse either above or below

the central base line of the display.

45
REFERENCES
4.1. D. G. Luenberg, "Resolution of Mass Spectrometer Data."
IRL-1021, November 1964.

4.2. W. Reynolds, "The Use of a Logarithmic Amplifier in Data
Processing of Analog Signals." IRL-1017, April 1, 1965.

4.3. J. F. Gibbons and H. S. Horn, "A Circuit with Logarithmic

Transfer Response over Nine Decades." IEE Transactions on
Circuit Theory, CI-11 (September 1964).

46
V. Computer Manipulation of Chemical Hypotheses

While the high resolution mass spectrometer is perhaps the most
capable single instrument for organic structural analysis, the sheer
volume of its signal output poses formidable problems of data reduction
and data analysis. These problems would be multiplied by the number of
samples that would need to be processed by the micro-scanning mass
spectrometer which is our ultimate goal. At one level, the problem is
the identification of mass numbers with compositional formulas. However,
no mass spectral signal is free of noise and great effort must then be
spent to obtain an accurate determination of mass to ultimate resolution.
Much of this effort is wasted when it does not answer a concrete question,
i.e., which of a set of possible compositions is indicated by a given
measurement. Even for all compositions, the corresponding mass numbers
are not continuously distributed; they are rather the discrete set of
numbers calculated from linear integral sums of nuclidic masses, and
represented in the tables.

The tabulations and calculation programs (references 5.1, 2,3,
4, 5, 6, 7, 8) are the first step in a control program for the mass
spectrometer. As soon as the peak is identified within a given mass
neighborhood, the competing possibilities should be computed, then
weighted in accordance with any other available information. This
allows the experimental problem to be restated as a choice among competing
possibilities, and the signal information need be accumulated only long

enough to lead to a meaningful choice among them.

47
In solving a structure, the chemist hypothesizes a series of trial
structures, then matches them with the data (in this case a mass
spectrum, but this can be generalized to any data set) and accepts or
rejects his trial solutions, usually part by part, in a structure.

Much of this tedious effort could be emulated or at least assisted by
the computer in a program we call "mechanized induction".

For this purpose, a language has been devised for representing
chemical structures in easily computable form; "Dendral '64"
(reference 5.2 and 5.5). The development of this ianguage required the
filling of a surprising gap; the systematic application of simple
topological principles to the field of chemical graphs - that is, a
symbolic representation of organic molecules. Existing notations were
found to be quite defective as the chemist already knows too well from
his difficulties with nomenclature (other organizations like Chemical
Abstract Service also recognize the problem and are working on it, but
tend to compromise topological rigor for the benefit of established
traditions in notation). At any rate, with the help of some theorems
on canonical forms of trees, and on Hamilton circuits of planar maps
(for acyclic and cyclic structures respectively), a complete system has
been worked out. This gives an algorithm by which the computer can
generate an exact list of all isomers in a given composition.

By itself this is a futile approach to any but the simplest
problems, since the number of possible isomers quickly exceeds the

range of a fast computer. Heuristic and symbiotic methods are therefore

48
called for whereby the computer emulates or cooperates in the use of
human problem solving techniques in searching wisely selected parts of
the space of possible solutians. Professor Edward Feigenbaum and Dr.
Richard Watson of the Computer Science Department participated in a
cooperative effort to program efficient displays of structural ideas for
conversational interaction with the computer. This was a step to
evaluate the chemists problem solving heuristics incorporating them in
the machine program. The effort was limited at present by the existing
computer facilities (a PDP-1 machine) with inadequate displays.

This work described above is covered in detail in the references
indicated and the additional items in the bibliography below. A complete
description of the current status is given in an article by J. Lederberg
and E. A. Feigenbaum?" ®, The following is the abstract of that paper.

"A computer program for formulating hypotheses in the area of

organic chemistry is described from two standpoints: artificial

intelligence and organic chemistry. The Dendral Algorithm for
uniquely representing and ordering chemical structures defines

the hypothesis-space; but heuristic search through the space is

necessary because of its size. Both the algorithm and the

heuristics are described explicitly but without reference to

the LISP code in which these mechanisms are programmed. Within

the program some use has been made of man-machine interaction,

pattern recognition, learning, and tree-pruning heuristics as

well as chemical heuristics which allow the program to focus

49
5.1.

5.2.

5.3.

5.4.

5.9.

5.6.

2.7.

5.8.

its attention on a subproblem and to rank the hypotheses in
order of plausibility. The current performance of the program
is illustrated with selected examples of actual computer output
showing both its algorithmic and heuristic aspects. In addition

some of the more important planned modifications are discussed."

REFERENCES

J. Lederberg, "Systematics of Organic Molecules, Graph Topology,
Hamilton Circuits." IRL 1040.

J. Lederberg, "DENDRAL~64. A System for Computer Construction,
Enumeration and Notation of Organic Molecules as Tree Structures
and Cyclic Graphs. Part 1." NASA CR-57029. STAR No. N65-13158,
IRL-1036, 1964.

J. Lederberg, "Tables and an Algorithm for Calculating Functional
Groups of Organic Molecules in High Resolution in Mass Spectrom-
etry." NASA - STAR No. N64-21426, IRL-1019, 1964.

J. Lederberg and M. Wightman, "A Subalgol Program for Calculation

of Molecular Compositional Formulas from Mass Spectral Data."
NASA - STAR No. N66, 11689, 1964. ,

J. Lederberg, "Topological Mapping of Organic Molecules."
Proc. Nat. Acad. Sci. U.S. 53: 134-139 (1965),

J. Lederberg, "Calculation of Molecular Formulas in High-Resolution
Mass Spectrometry." An appendix to Structure Elucidation of
Natural Products by Mass Spectrometry (Volume 11: Steroids,
terpenoids, sugars, and miscellaneous classes), by H. Budzikiewicz,
C. Djerassi and D. H. Williams. Holden-Day, Inc., San Francisco,
1964.

J. Lederberg and M. Wightman, "Calculation of Upper Limit of Hydro-
gens of an Organic Formula for Analysis of Mass Spectra." Anal.
Chem. Soc. 36: 2365 (1964).

J. Lederberg and E. A. Feigenbaum, "Mechanization of Inductive

Inference in Organic Chemistry", Symposium on Cognition, Carnegie
Institute, John Wiley & Sons, in press (1967).

50