Computers in the lipid research laboratory

Computers in the lipid research laboratory

Pro#. Liphl Re,~. Vol. 17. pp. 93 II0. Pergamon Press. 1978. Printed in Greal Brilain COMPUTERS IN THE LIPID RESEARCH LABORATORY R. O . BUTTERFI...

1MB Sizes 0 Downloads 68 Views

Pro#. Liphl Re,~. Vol. 17. pp. 93 II0. Pergamon Press. 1978. Printed in Greal Brilain

COMPUTERS

IN THE

LIPID

RESEARCH

LABORATORY

R. O . BUTTERFIELD, W . K . ROI-IWEDDER, E. D. BITNER, J. O . ERNST, D. J. WOLF,

and H. J. DUTTON Northern Regional Research Center, Federal Research, Science and Education Administration, U.S. Departnlent of Agriculture, Peoria, Illinois 61604, U.S.A.

I. GENERALCONSIDERATIONS

93

II. APPLICATIONS A. Chromatography B. Mass spectrometry I. Mass spectra 2. Data acquisition 3. Data processing 4. Selected ion monitoring C. Pilot plant D. Other applications E. Batch processing

95 95 97 97 98 98 101 102 105 107

111. STAFFING AND MANAGEMENT

108

IV EVALUATION V. REFERENCES

109 109 I. G E N E R A L C O N S I D E R A T I O N S

To computerize or not to computerize: "that is the question: Whether 'tis nobler to bear those ills we have than fly to others that we know not of?"

For his decision-making process, today's chemist is provided a more systematic approach for his soliloquizing than was Hamlet. The problem of automating laboratory instrumentation can be separated into well-defined tasks and executed in logical sequence. The procedure, as described in detail by Frazer ts and Perone et al., al consists of the following four steps: (1) project definition, (2) systems specifications, (3) function design, and (4) implementation. Whether intentionally or subconsciously these steps w~ll inevitably be followed in the course of implementing a successful plan of laboratory automation. In many instances, the chemist may be relieved of certain steps of this approach to computer installation. For example, it may suffice to carry out the first two steps by meticulously specifying service requirements and then allowing prospective vendors to propose solutions to the automation project, using their hardware to achieve maximum performance hopefully at minimum cost. In defining the system, all chemists can agree upon certain general requirements for the computer in their laboratory. These are: (1) Isolation of programs--so that each scientist is concerned only about his and not about other users' programs. (2) Accessibility--each user wants "his very own computer" and does not want to schedule his runs far in advance. (3) Real-time data acquisition and processing--the data must not only be acquired while the experiment is in progress but must also be rapidly processed and presented to the scientist in usable report form. (4) Control--must be provided to enable the scientist to modify procedures and control processing. (5) Ease of operation--is mandatory for the scientist to keep system programming and maintenance to a minimum. (6) Efficient operation of the system--should take into account the low duty cycle of some analytical instruments and share computer resources among users. (7) Reduction in cost--should be achieved under the concept of total laboratory automation in that individual expensive peripherals like bulk data-storage devices and highspeed printers and plotters can be shared by all scientists. The central premise from which these requirements spring and which underlies the hardware, software, and staffing of the general laboratory automation computer is that this computer tool is of, by, and for a practicing bench chemist. The chemist is king! 93

94

R.O. Butterfield et

~d.

As the field of laboratory automation has matured, there has appeared a viable alternative to total laboratory automation. This is the automating of individual or groups of like instruments by a vendor-supplied computer data acquisition system. These "turnkey" systems have several advantages. They are isolated, accessible, generally easy to operate, require no computer staff, and with the decreasing cost of computers, they are becoming cost effective. However, turn-key systems are not yet available for all laboratory instruments. They also have the disadvantages that their limited size sometimes prohibits simultaneous data acquisition and retrieval, that they generally come with slow-speed printing devices and limited data storage, and that they cannot be easily changed or modified to keep up with state-of-the-art improvements. C o m p u t e r equipment can be purchased to provide programming capabilities and increased data storage, but this increases cost. Either a scientist has to become a computer p r o g r a m m e r or a p r o g r a m m e r must be added to the staff. Thus, most advantages of the turn-key system are eliminated. In a large research organization, a centralized laboratory computer system can be the best alternative. It has the flexibility to meet the changing requirements of the scientist. With present-day operating systems, computer resources are dynamically allocated and users are isolated and protected from each other, giving them the impression that they have their own computer. Whereas computers have continued to reduce in price, high-speed printers and large bulk storage have not. On a laboratory-wide system, these devices can be provided and shared by all users, when no one user could justify the expense. Remote, stripped down, low-cost computers or microprocessors can be used where high data acquisition speeds are necessary, such as in mass spectrometry. These remote computers, with a minimum of equipment, can be totally supported by the large machine. This hierarchy of computers and sharing of expensive peripherals reduces capital expenses and results in a lower unit cost. Disadvantages of the total laboratory computer system are that its initial cost is high, it is not easy to implement, and it requires time to implement. A staff which has chemical, instrumental, computer, and electronics expertise is required. However, the resulting customized system can best do what the scientist wants rather than what someone else thinks he wants. At the Northern Regional Research Center (NRRC), we have taken the total laboratory automation approach partly because other alternatives were not available when we started and partly because our experience verifies that it offers the best service to the bench chemist at the lowest unit cost. Historically, N R R C started using digital computers in the mid-sixties by renting computer time at a nearby academic institution. In late 1966, we purchased a small batch type computer which we still use. Nearly a third of our professional staff was introduced to computer programming through in-house Fortran courses. During the late sixties, laboratory instruments had equipment added to punch paper tape for data logging. This tape was hand carried to and read by the batch computer which processed the data and produced tabulated and graphic output. In 1970, we acquired our first real-time process computer. We initially connected one mass spectrometer data system and 14 gas chromatography channels. By 1974, we had added another mass spectrometer, 16 more gas chromatograph channels, a scintillation counter, and were acquiring data from a laboratory-scale soybean oil refinery. To take advantage of reduced computer costs and a superior operating system, in the summer of 1974 we converted to our second real-time process computer.* Since then we have added a third mass spectrometer, 24 more chromatograph channels, including amino acid analyzers, liquid chromatographs, and a plate scanning densitometer. Figure 1 is a diagram of this system. We plan to add an additional 40 channels of signal inputs and a fourth mass spectrometer in the near future. A detailed description of these applications follows. * For information, the computers referred to in sequence are: IBM 1130, IBM 1800 and Modcomp II/25. Mention of firm names or trade products does not imply that they are endorsed or recommended by the U.S. Department of Agriculture over other firms or similar products not mentioned.

Computers in the lipid research laboratory

Remote Laboratory Locations

Computer Room

[ Console ~\ I PaperTapeJ ~/~

I[150OisStora, h L\'q e ,,'/!,,Jl a lei[a Oytes

MaeneticTape " Drives

,

NuclideMass SpecI

PDP12Computer H

FinniganMass Spec]

[~.-.~Dupont Mass Spec]

Nova 1210 Computer I -'1 16 GC Channels

I

10 GCChannels I

ScintillationCounter I

Gilford Speciophotometer]

Printer/Plotter

XY Plotter

POPB/I Computer ]~

A/D Converter

I Electrostatic [

95

a

tier i~'~SoybeanOil Mini-RefineryJ

I Key Punch/ Card Punch [ FIG. I. Diagram of NRRC's computer system.

I1. A P P L I C A T I O N S

A. Chromatooraphy The gas chromatograph (GC) is one of the most c o m m o n laboratory instruments and can be easily automated. The cost savings will not be as great as automating a mass spectrometer; however, such automation can pay for itself in several years along with improving accuracy. Problem definition seems simple. There is a time dependent signal composed of normal distributions which one wishes to integrate. However, let us take a closer look at the problem. Gas chromatographs differ in many ways and there are many types within our building. Some machines may have thermal conductivity detectors which produce millivolt signals with a usable dynamic range of 1000-1. Ionization type detectors can produce volt level signals with a usable dynamic range of 100,000-I. Peak width can vary from seconds wide to minutes wide so that sampling rates have to be variable in order to have an adequate number of data points across a peak. Some sort of control switches and indicators must be provided so that the scientist can tell the computer when to start and stop data acquisition and to inform him as to the computer status and how it is working. Once the data are acquired, a computer program must detect peaks, correct for baseline drift, resolve overlapping peaks, and produce a report of the results. At NRRC, three data acquisition systems for GC's have evolved. The first consisted of a custom-made, remote analog to digital converter and multiplexer around which eight GC's were grouped. The unit transmitted digital signals to the central computer. A second group of GC's were concentrated around a minicomputer which acquired data from a mass spectrometer as well as from the GC's and again transmitted the data digitally to the central computer. Both these systems were built at a time when it was felt that one could not transmit low-level analog signals for the required distance. This approach required the grouping of GC's and did not provide service to the isolated machines. To service all chromatographs, we purchased an autoranging, multiplexing analog-to-digital input device with our most recent computer. It has 12 voltage ranges from _+5 mV to _+ 10.24V and can be expanded to accept 256 inputs. At the remote G C site, a control box as pictured in Fig. 2 is provided. This in-house built box serves the three functions of control, status display, and signal conditioning.

96

R.O. Butterfietd et al.

FtG. 2. Gas chromatograph computer input cont.rol box.

It consists of three indicating switches and a high impedance, differential input amplifier. The high impedance allows us to connect to any circuit without loading. Differential input permits us to choose our signal between any two points rather than between one point and ground. The amplifier also can have a fixed gain so that low-level signals are amplified. Three indicating switches are used by the operator to start and stop data acquisition. They tell him when the computer is ready to acquire data, that data are being acquired (start light on), and that acquisition has stopped and the data are being processed (stop light on). When the processing is done, the stop light goes off and the next run can be started. No matter how raw data is acquired, the same GC processing program is used to detect peaks, correct for baseline, and generate a report. The scanning logarithm used is a slightly modified version of IBM's McCullough Gas Chromatograph Monitoring Program. 2s First and second derivatives of the raw data determined when peaks start and stop. After peaks are detected, a second program corrects areas for baseline and types a report on a teletype at the remote location. Figure 3 shows such a report. This preliminary report consists of sample documentation, retention time, area, and percent area. The purpose of this report is to advise the operator that his data have been acquired, whether the sample injection was adequate or not, and that the resolution of peaks are as expected. In many cases this is all the chemist wants. In some routine applications the computer, as directed by the scientist through use of a remote teletype, further processes the data to produce an identified report with mole 9o and/or other

CH 90 MTD 74 RUN 121 7 7 77 CU HYDROGENATED SOYBEAN OIL VARIAN 1800-FID ME ESTER ANALYSIS 15%EGSS-X.CHROMOSORB W 100-120 MESH ISOTHERMAL 185C PK

TIME MIN

RELATIVE AREA

RELATIVE PCT

1 2 3 4 5 6

4.89 832 9.90 12.72 17.23 20.30

830564. 323998. 2581273 4255244. 199309. 51142.

10.08 3.93 31.32 5163 2.42 062

TOTAL

835

5FT-1 8

TEMP C

8241530.

FIG, 3. Gas chromatograph preliminary data report. Peaks are: (1) palmitate, (2) stearate, (3) oleate, (4) linoleate, (5) linolenate, and (6} conjugated diene.

Computers in the lipid research laboratory

97

units the scientist desires. These reports can be given to the person who requested the analysis and directly entered into the scientist's notebook. Although our discussion has centered on gas chromatography, other types of chromatography can also be automated by the same hardware and software. We routinely acquire data from a densitometer used to scan thin-layer chromatographic plates. We have two amino acid analyzers on line and several high-performance liquid chromatographs. These devices benefit as much as GC. Computerization of chromatographic equipment has many advantages. Accuracy is increased, reproducibility is improved, and labor is saved. But, as Cram and Juvet have observed: "The ultimate power of a data system to the chromatographer lies in the software. ''9 It can just integrate and report as some systems do or it can control the instrument, use nonlinear least square curve fitting for resolving highly skewed peaks, t° calculate retention indices, plus convert areas to whatever unit is meaningful to the scientist. Data can be stored and collated with other results for given samples. The only limits are one's needs and one's imagination.

B. Mass Spectrometry The combination of a gas chromatograph and a mass spectrometer (GC-MS) produces a system which is capable of generating a complete mass spectrum every 6 sec, l0 spectra per min, or 600 spectra per hr. This presents the scientist with two problems: How does he collect 600 spectra in an hour, and once he has them how does he manage to look at and understand that much data? The computer is the answer to both problems. Over the years mass spectrometers have been a challenge to the state-of-the-art of data recording. Recording methods have included strip chart recorders, photographs of oscilloscope screens, light beam oscillographs, punched paper tape, magnetic tape, and hardware digitizers. Both punched paper tape and magnetic tape have been used to hold data to be processed on a remote computer. As the price of computers has decreased more and more, they have been used to collect data directly until now they are no longer a luxury but have become a necessity. The purpose of this section is to give the lipid chemist an idea of the problems of connecting a mass spectrometer to a computer and also a suggestion of the many things the computer can do for the chemist. Emphasis has been placed primarily on G C - M S , since it presents the greatest problems in both speed and quantity of data. Computers are also important in selected ion monitoring and high resolution work.

1. Mass Spectra A typical mass peak for a fast scan is shown in Fig. 4. Smoothness of the curve depends on the number of ions being measured, and the peak may break up into several parts for very small peaks with few ions. Two characteristics of the peak are important, the intensity or area under the curve and the position or mass of the peak. G C - M S requires a very wide dynamic range to cover both the wide variation in sample sizes of the GC eluate and the span of mass peak intensities. A dynamic range of 100,000:'l is hardly sufficient. Some mass spectrometers are equipped with 12 bit analogto-digital converters (ADC) which have a dynamic range of less than 4096 (depending on what is considered an acceptable accuracy for the smallest peak). Sixteen bit ADC's have a dynamic range of less than 65,536, but they are very expensive in the speed range required by mass spectrometers. Autoranging 12 bit ADC's are a compromise between the factors of speed, accuracy, and cost. Hysteresis and inductance effects of the magnet require that magnetic mass spectrometers be scanned in one smooth continuous sweep through the entire mass range. Data are collected by making an analog-to-digital conversion every 100/~sec or less as shown in Fig. 4. If a very strong peak causes the ADC to go over range, that

98

R . O . Butterfield et al.

I00 Microsecond Intervals ot

Analol to nilital Conversions

FiG. 4. Single G C - M S mass peak.

peak is lost and the whole spectrum is distorted during normalization. Quadrupole mass spectrometers on the other hand use voltage scanning and can be started and stopped at every mass so that, in case of an over range, the same peak can be measured again at lower amplification. This ability to remeasure a mass at lower gain greatly increases the effective dynamic range of the quadrupole mass spectrometers over magnetic instruments. Attainable accuracy of G C - M S spectra is limited by the continuously changing intensity of the G C peaks and the small sample size. However, accuracy is important in isotope and quantitative work, where the nature of the sample and the mass peaks of interest are well known. Mass scans are usually very slow or discontinuously stepped from mass to mass. Accuracy can be increased by lengthening the time constants of output amplifiers so as to integrate the ion beam to reduce ion statistics problems. The 12 bit ADC's can be software integrated for 1000 or more conversions, or slow but high-accuracy digital voltmeters can be used to record the peak intensities. The other important characteristic of the peak is the mass which must be measured to sufficient accuracy to provide nominal mass marking over the whole range of the mass spectrometer. Nothing can destroy the value of a mass spectrum or a mass spectrometer faster than mass marking errors. Mass is usually measured as a function of magnetic field (Gauss), time, or voltage. Magnetic mass spectrometers use a solid-state Hall effect device inserted between the poles of the magnet to measure the magnetic field. The voltage from the Hall device must be digitized with at least a 14 bit A D C to get sufficient accuracy. Time can also be used as a mass base if the ADC is driven by a precise clock and the magnet scan is controlled with a good quality RC circuit. In this case, every analog-to-digital conversion is considered a clock tick used to increment a counter in the computer, which is the mass base. Both Gauss value and the time base (with linear scan) must be squared to be proportional to mass. A mass marking system using Gauss requires more equipment but is preferred to a time base system which requires frequent calibration. With the quadrupole mass spectrometer, mass is directly proportional to scan voltage and the computer can compute the voltage for each mass and directly control the quadrupole mass scan.

2. Data Acquisition A mass spectrometer peak as measured during a typical G C - M S run is shown in Fig. 4. The ADC makes from 10 to 20 measurements of intensity across the peak in 100#sec intervals. This results in more than 10,000 measurements per sec which

Computers in the lipid research laboratory

99

I00

riO

a~

"i 60 -= ,~o

I

20 J l,

20

'l

l)

40

l"

60

Ill

l

'

I

"

,

,

,

"

I

"

"

'

"

I

IO0 120 140 160 IIIO 200 220 240 260 200 300 320 340 Mass Number

FIG. 5. Mass/intensity plot of methyl stearate (mass spectrum).

must be handled by some device, usually a dedicated minicomputer. A time-shared or multiuse computer will not do, because it would be overwhelmed if it tried to measure intensity every 100/~sec and serve other users at the same time. The dedicated minicomputer would be overwhelmed if it tried to store 10,000 measurements per sec or 60,000 per mass spectrum for 6-sec scans. The bulk of these 60,000 measurements are worthless baseline measurements with only a few hundred groups of measurements representing peaks. To keep down the total amount of data, the dedicated minicomputer processes the analog-to-digital conversions in real-time, as fast as it is being generated, throwing away the baseline data and adding the readings together to form single values for each peak. At the same time, the computer must also handle the time or Gauss readings, finding the center of the peak and attaching the correct time or Gauss value to the intensity value. Thus, the 60,000 measurements taken in a 6-sec scan are reduced to a few hundred pairs of computer words representing the masses and intensities of the mass spectrum. Several hundred of these scans represent a G C - M S run. Most dedicated computers do not have enough main memory to store this processed data and must store it on disk. Magnetic tape can be used, but it is inconvenient for on-line data collection. An excellent and efficient technique is to let the dedicated minicomputer compress the data and pass the condensed spectra and the pairs of mass and intensity to a multiuse computer for processing and storage. Quadrupole mass spectrometers can step from the center of one mass peak to the center of the next mass peak, thus reducing the time wasted measuring baseline and greatly reducing the amount of raw data collected. Care must be taken to be sure that the top of the peak is measured, and the mass defect of each of the peaks must be taken into account.

3. Data Processing The most important things the mass spectroscopist wants from his data system are mass intensity plots, as shown in Fig. 5. Before the use of computers, it took a good part of a day to mass mark, normalize, and plot a spectrum; but now, one expects the computer to do it in a minute or two. The G C - M S computer can produce far more spectra than any scientist can possibly want. To reduce this data, it is first necessary for the computer to sum the intensities of all the peaks in each mass spectrum and to plot the sum for each scan as a single point against scan number. 22 Scan number is a direct function of time, and the result is a total ionization plot which is very similar to the G C curve. The total ion plot provides a visual index of the accumulated data, and the scientist can pick out the G C peaks of interest for more careful analysis. J.P.C.F. 1 7 1 - - G

1(I~

R.O. Butterfieldet

al,

Total Z

J

Mass

, Mass

io

2o

ao

4o

so

io

io

ao

FIo. 6. GC-MS total ionizationand mass chromatographyplots of an aldehyde-aldehydeester mixture. A total ion plot of a mixture of aldehydes and aldehyde esters from the oxidative cleavage of methyl octadecenoate with its double bond scattered during catalytic hydrogenation is shown in Fig. 6. The figure is a computer-expanded view of 80 spectra out of the center of a GC-MS run consisting of over 300 spectra. The mass spectra were scanned at a rate of five scans per min, which can be seen to be inadequate because each GC peak contains only one mass scan. To be assured of getting a go6d mass spectrum, there should be at least three or four scans covering the top 80% of the GC peak. This would require 15 scans per rain or 900 scans to cover the whole GC-MS run. Also included in Fig. 6 are mass chromatograms 2a for masses 44 and 74 which are characteristic peaks for aldehydes and methyl esters, respectively. These curves are generated by having the computer search through the entire GC-MS file for the intensity values of masses 44 and 74 and by plotting them against scan number. The plot of mass 44 in Fig. 6 highlights the positions of the aldehyde peaks while the plot of mass 74 highlights GC peaks containing methyl ester functions. Something new was discovered in this GC-MS run: the mass 74 line also has maxima under, but not aligned with the aldehyde peaks. The aldehyde GC peaks which have been used for quantitative calculations have normal methyl esters eluted with them. Mass chromatography very easily showed these methyl ester peaks, which would have been very difficult to detect otherwise. The computer can reprocess the GC-MS data in many ways including background subtraction, peak summation, and mass spectra library searching. Most GC-MS spectra contain extraneous background peaks due to water, air, and column bleed which can be subtracted from the measured peak. Depending on the nature of the sample and tile background, useful results can be obtained even if the background is ten times the intensity of the sample. Noise can be reduced and sensitivity increased by summing mass intensities over a GC peak. Peak summing can also be used for quantitative work, although selected ion monitoring is a superior procedure. (See below). One means of identification of a mass spectrum is direct comparison with known library spectra. This has been done manually for many years using the eight largest peak files. 1+ Tbei'e are many library-searching routines in use and one of the best is

Computers in the lipid research laboratory

I01

that of Hertz et al. 21 This program reduces both the library spectra and the unknown spectra into mass and intensity values of the two most intense peaks in each 14 mass units. If the base peak in each spectrum is given the value of 63 and the rest of each spectrum is normalized accordingly, both the mass and intensity can be fitted into one 16 bit computer word and the whole 36,000 spectra of the National Institutes of Health (NIH) file can be fitted into 4 million words of disk memory. With the file on disk, the scientist can type in the identification number of his spectra and get an automatic search of the 36,000 library spectra. The Hertz et al. system uses a similarity index which varies from 1 for a perfect match to 0 for a very poor match. Some form of limited library search routine is usually provided with the newer G C - M S computer systems, but they do not usually have enough disk memory for the full NIH file. Computer matching and interpretive systems are available over telephone lines if the user has the proper terminal. Cornell University supports two systems, the SelfTraining Interpretive and Retrieval System (STIRS) 26 and the Probability Based Matching system (PBM), 32 which are available over the T Y M N E T computer network. Another system is the Chemical Information System (CIS) 2° operated by the National Institutes of Health and the Environmental Protection Agency. 4. Selected Ion Monitoring

G C - M S does not make very efficient use of sample, particularly if one ion alone can be used to characterize the sample. G C - M S scans the whole 400 or more mass numbers of the spectrum and wastes the time it takes for the mass spectrometer to return from the highest mass back to the starting mass. Also, the mass spectrometer spends more than half of its time looking at the baseline between peaks. For quantitative work where the identity and mass spectrum of the sample is known, the mass spectrometer can be stepped from one mass of interest to the next, ignoring most of the spectrum and all of the baseline. This is a very old technique a4 more recently given the names selected ion monitoring (SIM), multiple ion monitoring, or mass fragmentrometry. It provides more than a 1000-fold improvement in accuracy and sensitivity over G C - M S . While SIM can be used to follow a single ion of a single compound, better quantitation can be achieved if an internal standard, ofter an isotopic isomer of the compound of interest, is included with the sample. The computer controls the mass spectrometer during selected ion monitoring, setting the voltage to the desired mass, timing data collection, software integrating the analogto-digital conversions, and automatically outputing the final result as a percentage. Figure 7 is a SIM MS curve of methyl esters of phosphatidylcholine fatty acids from blood of a human subject who was fed deuterium-labeled oleate and elaidate. 17 The SIM MS followed the four components, methyl oleate-do, methyl oleate-d2, methyl elaidate-d4, and methyl stearate, by monitoring the M-32 (molecular ion minus C H a O H ) region, mass 264 through 269 as they were eluted from an Apiezon L G C column. All the operator has to do is type in the sample number and title and inject the sample into the GC. The computer controls the mass spectrometer, repeatedly scanning through the six mass peaks, setting the mass for each peak, collecting the data for 100msec, and continuing on for the duration of the run. After completion of the GC run, the computer automatically calculates the total ion curve shown in Fig. 7, finds the front and back of the oleate peak, scans 221 and 590 in Fig. 7, and sums all the intensity values under the curve for each mass. It finds the lowest point in front of the peak and sums 50 scans on each side to form a background value which is subtracted proportionally from the peak value. Still automatically, the computer retrieves the values from scans made from pure methyl oleate, methyl oleate-d2, and methyl elaidate-d4 and forms simultaneous equations with them and the unknown values. It solves these equations using an iterative least squares method to find the relative amounts of each of the standards which make up the unknown blood lipid sample. The computer recalculates

102

R.O. Butterfield et al.

100 9O Total Ion 80

7oi 60

50l)leate Cut • 221 to 590

,ol i/ ll/ 3O

Old2 M.32]

20110

221,

0

"

0

1

2

"

/

-

' . . . . .

3

4

J

.

5

.

.

.

.

6 7 Spectra/lO0

.

.

.

8

.

-

9

-

I0

11

12

FIG. 7. Selected ion monitoring plots of methyl esters from human blood lipids.

the values to give the percentage of labeled material in total fatty esters, and the percentage of elaidate in the label (ratio of elaidate-d4 to oleate-d2) to give the final result. To make optimum use of the computer, the operator types in starting and stopping times just before he injects the sample, and the computer starts and stops data collection at the proper times while the operator goes to lunch. Thus, the computer has not only made the work possible but it has become the perfect technician, e.g. it never tires, never has to leave the room, never makes a mistake, and does exactly what it is told to do, right or wrong. C. Pilot Plant

The conventional application of the on-line computer is process control. This is exemplified by refining in the petroleum industry, which may be considered to be all automated and largely under computer control. Of course, the cracking or distilling process of the petroleum industry is much simpler than the stages of processing in the vegetable oil industry. 33'a6 Due to the complexity of vegetable oil refining, a "turn-key" type computerization is not yet available and a "user developed" or "in-house" system must be implemented. Thus, a high price tag is inevitable and other factors for computerization must be stressed, such as the ecomomic advantages in uniformity of product, improved efficiency of operations, and minimization of human error. Strange as it may seem to the scientist, simply providing continuously updated production information may be the computer's strongest selling point in industry for management. Present indications are that automation, with its accompanying computerization of v.egetable oil refineries, is being initiated for one stage of processing at a time. s'tt'lS'16 A major problem that had delayed computer control (automation) of vegetable oil refining has been the virtual absence of transducer devices that convert chemical physicaL or mechanical phenomena into electrical signals. A method developing and testing transducers was divised at NRRC, USDA, Peoria, Illinois, and consisted of a miniature soybean oil refinery (minirefinery). Figure 8 is a pictorial schematic of the minirefinery. The four basic operations which are similar to industrial plants are refining, bleaching, hydrogenating, and deodorizing. This arrangement permits the observation of trans-

~Samp LrJ Size

[

PumPlDrylce[~

v,c ~

.~2 ,~

Deodorize

, Samp

~"--H2O ~

Pt

i7

0.rI

==

==

~

Waste .

~

~ L~

h

e2

i

Dry

~

Hydrogenate

I ~ '~II~eP~L_ J. Filter\,.).. v~-,-

H.t,r , . ~

v

Mazouts"----'--.~r'-Ta°h

~ ~ Waste - - *

.,0

~oom

Fl(3. 8. Miniature soybean oil refinery process diagram.

DistillationColumn

I-

T'C .I,.... I. Deaerate I

Heater

*

Motor

i:

Wash

[- Motor

Neutralize

Stir

"LI I~

H,I

[

P1

~~ / /

Is,rap

I

I

V~c

I~...,,.~

Motor _1

X~

~-I-ii""'Heat

..,-~

t,rth

Filter o

O

c~

O

¢"3

104

R . O . Bulterfield et al.

ducers under conditions resembling an operational commercial facility. The complexities involved in automation and computerization will become obvious by studying Fig. 8 and later figures, especially when one realizes that industrial plants are even more involved with storage facilities, much greater capacities, and need for monitoring losses of any type Ioverflow, poor quality oil). The alkali refining portion of Fig. 8 includes the sections marked neutralize, wash, and dry. Thermocouples ITC), a tachometer (TACH), and a pH meter (pHt are the transducers in use in this stage. A sample of 0il for automated analyses is obtained at the point marked "SAMP.'" The bleaching operation consists of three steps: (1) mixing bleaching earth with oil, (2) bleaching the oil, and (3) filtering earth from oil. A thermocouple and a vacuum sensor are the transducers now in use in the bleaching step, but earth flow, oil flow, and others may be added. The hydrogenation stage consists of a heated column stirred by a rotor, with hydrogen and oil containing catalyst flowing through it. Many transducers are in use in this stage for following the parameters of the reaction and consist of (a) infrared (IR) instrument for indicating amount of t r a n s present, 12 (b) refractive index (RI) instrument for correlating with iodine value and amount of double bonds, 36 (c) G L C equipment for individual fatty acid composition, (d) a thermocouple reaction temperature monitor, (e) a stirring rate (rpm) tachometer, and (t) a hydrogen flow (ml) indicator for controlling the rate of hydrogenation. ~'36 The final stage in refining is deodorization. This is the most difficult stage to determine satisfactory completion with transducers, since odor and flavor of finished oil as determined by the human senses of taste and smell are the two most important criteria. At the present time, the three parameters that have proven critical to deodorization are temperature, steam flow, and vacuum, so these are monitored by transducers. Oil flow through the four refinery stages depicted in Fig. 8 can be traced by following the darkened line. A more detailed description of the minirefinery and industrial operations can be found in the literature. 4"33'36 Additional transducers are shown in Fig. 9, which is a simplified diagram of the automated analysis apparatus. It is used to analyze the oil samples taken from the minirefinery from the points marked "SAMP" in Fig. 8. Signals for the computer are obtained from the color photometer, sodium flame detector, free fatty acid (FFA) titration, and six-way (6W) valve. A sample is analyzed by being pumped from the refinery (start with degummed--*) through a solenoid (SOL), the 6W valve, a sample directing SOL, and to either the color photometer and sodium detector in series or to the FFA titrator vessel. The transducers supply the computer with signals representing the sample's origin (position of the 6W valve), the color of the oil, the sodium content of the oil, the weight of the oil sample for titrating, and the amount of alkali used for titrating. A more detailed description of Fig. 9 can be found in the literature. 4 Figure 10 shows a simplified block diagram of the minirefinery operation. The only blocks on this figure that do not actually involve some electronics are the "Refinery" and "Samples" blocks. Signals from the transducers (Xsducers) of the process and automated analyses arrive at the computer by routing through the recorder multiplexer, digitizer, and interface. In turn, signals from the computer (dashed lines) can, if desired, control certain processes (DDC), timing (system controlled), or printout (TTY). In addition, the teletype can be used to enter additional information or to access other programs. At this stage of development, the minirefinery is one example of laboratory computerization and exemplifies possible industrial application. Many modifications, additions, and deletions are continually occurring as testing and evaluating continue. Other transducers, such as iron, phosphorus, and moisture detectors are under consideration. Changes in computer systems, multiplexers, and other wiring are being implemented. Of course, in the laboratory as in industry, new and improved technology in computers, such as microprocessors, will continue to change the hardware and improve the performance/cost ratio.

Computers in the lipid research laboratory

105 To Hood

/--.



T'o Rec. _.]'~ Flame ~ or Mux I_~_lDetoctor I -

1

-

[

~

- -

I Strainl _ To Kec.

I

y

l Hydrozenated

111" !3J

[Soil

~

Sol P Rec Mux MEK •

~, Blea~:hed

Solenoid Pump Recorder Multiplexer Methyl Ethyl Ketene To Refinery

FIG. 9. Diagram of automated analysis apparati for miniature soybean refinery.

D. Other Applications Scintillation counting is an application in which the computer can relieve a chemist of tiring calculations; however, the volume of data is not sufficiently large to justify a computer system by itself. A benefit of total laboratory automation is that such an instrument can be placed on-line since only the cost of connection must be justified. In most cases, this cost is small since modern day scintillation counters type their data on teletypes, and by using standard computer components, this signal can be read by the computer. The character string is decoded into sample number, counts per channel, and external standard ratio.

Refinery

I.E.. . . . . .

FIG. 10. Block diagram of soybean oil refinery components.

106

R . O . Bultertield et al.

1.0 0.9i

[]

0.8 0.7

~r', ~ 10~

° ''''''--~°

°'u

_g o 06 "'=" 0.5

~'~

0.1

~ o

0



---~""TJ" 0.1

. . . .

0.2 0.3 0.4 0.5 External Standard Ratio

0.6

FIG. 11. Scintillation counter calibration curves. × Tritium channel A, Q Tritium channel C, /x Carbon-14 channel A, • Carbon-14 channel B, [] Carbon-14 channel C.

As discussed in detail by Thomas and Dutton, 37 the external standard ratio is related to the counting efficiencies in each channel for each isotope. Having counted a series of quench standards, the computer fits a cubic function to the data, plots the data (Fig. 11) for a visual check, and remembers the determined constants. Reversing the process for an unknown, the computer calculates isotope channel efficiency from the external standard ratio and then convertscounts per minute to disintegrations per minute. For dual label work, simultaneous equations are solved in which the total count in a channel is the sum of that contributed by each isotope for each channel. This calculated data is then tabulated and/or plotted. Another advantage of total automation is that if the data is from the effluent of a chromatograph as described by Thomas and Dutton, 37 the inactive chromatographic curve can be collated with radioactive data and plotted together as shown in Fig. 12. 27 This computer plot immediately visualizes the fact that in the phospholipid portion of an egg, a hen elongates feed linoleate-l-t4C to arachidonate where the feed linoelaide-12(13)3H is not elongated. Thus, the chemist is relieved from tedious calculations and his data are presented to him in an immediately useful form. A Gilford spectrophotometer analyzer is in the same category as a scintillation counter. By itself may be difficult to justify automation; however, its digital output can so easily be connected to an already existing computer that this is a reason to connect it and to explore all the advantages inherent to computerization. Readings are converted directly to concentration units and tabulated along with original readings. The chemist has thus been saved the calculation and transcription of the data. It is immediately usable and transcription errors have been eliminated. Nuclear magnetic resonance spectroscopy (N.M.R.) has used computers extensively. Many computer programs exist for interpretation and calculation of N.M.R. spectra, for analyzing N.M.R. of polycrystals, for computing powder line shapes of quadrupoles nuclei, for deconvulation of broad-line N.M.R. spectra, etc. 39 Many new instruments include a computer which is required for pulsed Fourier transform N.M.R. Without the computer, 13C N.M.R. would not be a reality. N.M.R. is an application like mass spectrometry where a dedicated computer is required because the sampling rates are too high to leave the computer much time for other operations. The computer can

Computers in the lipid research laboratory

j l

107

,i

I

il;

.....

i11

II

Thermal Conductivity 3H Assay

....... "c A.ay

lis;

3

i llflA 4

lili Ji ti i. o. d ! ii ii

I~ :~ li li ! IJ!

Ii sI!

....

P li : :, t

1

r

l

6

12

18

~ i

"6 i "~

r

T

r

24 30 36 Retention Time, min.

T

T

42

48

.

54

FIG. 12. Computer drawn gas-liquid chromatography/liquid scintillation counting of methyl esters from egg yolk phospholipid fraction. Major peaks are (l) palmitate, (2) palmitoleate, (3) stearate, (4) oleate, (5) linoleate, and (6) arachidonate. be used to control the N.M.R as well as to acquire data and, because of this, a whole series of experiments can be performed unattended, t9 For example, an 8-hr long-term time-averaging, off-resonance, and Tt experiment can be executed automatically overnight. The chemist's productivity is definitely enhanced by use of the computer. Many laboratory instruments such as X-ray spectrometer, a infrared spectrometer, 2s surface area analyzers, and even balances can be automated. Reasons for computerization run from "it cannot be done any other way," to the need to reduce human error and to increase productivity. Whatever the justification, the chemist benefits.

E. Batch Processin9 The availability of the central computer along with its peripheral input/output and bulk storage devices allows one to establish a "batch" type operation. The initial function of a batch system is to allow the user to write necessary support programs. The usefulness of programs which are not related to the on-line tasks soon becomes obvious. Statistical calculations, such as regression, analysis of variance, etc. are effectively programmed for the computer. There are a myriad of uses to which the computer may be applied in scientific fields, z'24.29"as In one program, the digital computer replaces a function originally served by the analog computer, v'3° The computer is p r o g r a m m e d to read a set of data describing the amounts of various fatty esters present at different times during hydrogenation. The computer adjusts the theoretical rate constants until the calculated results match the laboratory data. The resultant calculated rate constants along with a plot of fatty esters concentration vs time or vs double bond provide the scientist with the results of how well his data matches his kinetic model. 6'as Another program allows the scientist to optimize the use of the counter current distribution or counter-double current distribution apparatus. In this program, the scientist can use different solvents, transfer volumes, and feed tubes to calculate a combination that will yield the desired separations and products. Different runs on the computer require little time and none of the tedious clean up involved when using the real equipment. Once a satisfactory set of conditions is determined, the scientist uses those conditions for one good run on the apparatus, s'~a

108

R.O. Butterfieldet

al.

Another set of programs which we have established is for our conversational mode terminal support. A push button at each terminal allows anyone to enter the name of a program he wishes to use. The programs presently available fall into the following categories: (1) control of data acquisition, (2) processing of acquired data, (3) scientific calculations, and (4) data base retrieval. An example of the latter is our literature file which contains a keyworded index of all NRRC-produced literature in the field of oilseeds since 1961. A scientist can search this data base by keyword, author, and/or phrase in the title and retrieve a complete citation list for those references meeting the search criterion. Again the chemist is aided in another phase of his work by the presence of a computer facility. !11. STAFFING AND MANAGEMENT The Lincolnesque paraphrase of computer philosophy "of the chemist, by the chemist, and for the chemist" should be reflected in the computing staff and organization. It follows therefrom that the general responsibility in a chemical laboratory for the computer project should lie in the hands of a chemist whose goal is making sure that the computer is available for use by chemists. At NRRC, positions in the computer staff consists of: a chemist kineticist who is the head of the group and is responsible for interfacing of real-time experiments; a formally trained chemist with a strong computer bent, showing even at the undergraduate level, whose prime responsibility is to keep the computer system running along with some applications programming; an industrial engineer, basically trained to solve problems, who has an advanced degree in computer science and is responsible for our continuing mass spectrometry program; and a part-time key punch operator as needed. This minimal and efficient staff contrasts sharply with that of the heavily staffed batch-oriented type computer installations. Its operation was predicated upon the premise that the scientist will be his own Fortran programmer and thereby eliminate the scientist-programmer communication problem. Under this concept, the computer staff guarantees to acquire data from laboratory instrumentation and store it on a logical file where it is available to the scientist to manipulate with his own program. During the initial implementating steps, the above premise and organizational concept helped get several instruments on line quickly. However, support for such a general instrument as GC and such a complex instrument as a mass spectrometer was written by the computer staff. Now that the basic acquisition programs are written and the system is running smoothly, the staff finds itself able to aid scientists in their programming, to do application programming upon request, to study the individual interfacing problems previously contracted to custom houses, and to plan for the inevitable changes in computer hardware which accompany the new electronic developments. While our computer installation was planned for whole building automation, the computer project in fact has grown up in one of our research groups whose main commodity responsibility is for soybean oil research. Here the need and motivation for the application of computers were present, The computer staff was drawn from chemists with strong interests and a desire to develop computer expertise. The computer project then developed under the efficiency of a hopefully benevolent dictatorship which recognized both the research project responsibility as well as the need for computer application. This greatly eased the problems of implementation and helped smooth the disturbance of change, especially during the initial learning phases. However, as the computer project has expanded its service to other research groups, it is only natural that administrative responsibility should reside with a more general authority of the research installation. The results of this philosophy of organization and staffin~ efficiency, responsiveness, and overall productivity are enhanced because the computer staff are scientists in their own right. They can contribute to each project and there are a minimum of communications problems since a common language is shared.

Computers in the lipid research laboratory

109

IV. E V A L U A T I O N

In the field of computers, there is only one thing that may be said with certainty and that is, what is taken for truth today will certainly not be true tomorrow. Because of the technical developments in electronics which result in lower cost computers, the distinction between the minicomputer and the large computer has all but disappeared and with it the sharpness of the argument between proponents of the minicomputer and the larger laboratory automation computer. Under the concept of hierarchical arrangement of computers, the argument appears to entirely resolve into a determination of what functions can most efficiently be achieved at which location by which computer. It is not so much a matter of whether a mini- or maxicomputer should be acquired but rather the planning ahead for their inevitable interconnection with mutual benefit that should be considered. Nor should the hierarchical arrangement of computers concept stop merely with laboratory computers. There is the need for the chemist working at his laboratory bench to call upon even large computers to carry out his simulations and to interrogate data banks such as the ACS Chemical Abstracts literature file or the Chemical Information System of the National Institutes of Health. 2° Management also has a stake in this hierarchical computer arrangement. Information on research projects, budgetary data, and research reports can be transmitted through, if not generated by, the computer. The moral to be drawn for the administrator is that he should interact with the computer programmer early in the process so that the computer becomes his tool and not his straight-jacket and so that it provides information for his decision making and does not attempt to dictate his course of action. So you decided to acquire a computer and did so. Now what has it done for you'? Assuredly, it has disrupted your life style, it has changed your whole approach to research, it has accelerated the pace of life. In place of the mountains of recorder paper which the computer was suppose to level, you now have erected mountains of computer output. You have spent a decade of your life and inverted the lives of your associates. You have inadvertently forced fellow chemists to learn a new language. Was it worth it? A glance at the publication list of NRRC for the last decade reveals the increasing influence of the computer until hardly a publication now does not feel its effect. This does not mean that all the staff are programming; rather, they are now exploiting the time savings of gas chromatographic automation, mass spectrometry automation, etc. Or sometimes the computer performs a statistical calculation for them that goes unnoticed and they incorporate computer printed reports into their notebooks without a second thought. The time savings anticipated for mass spectrometry have been even greater in practice than expected. A G C - M S run can produce as many as 600 spectra per hr. The mass of data recorded exceeds comprehension and yet each one of the spectra is available for inspection upon demand. In this new computer world, the magnetic storage of the computer becomes the scientist's laboratory notebook. Where we used to brag that "~the difficult we do right now and the impossible takes a little longer," now we are doing every day that which was impossible a decade ago. And even when the calculated savings do not run into the hundreds of thousands of dollars, the precision, accuracy, and improved format of this data handling recommends itself to today's chemist. As an aging farmer once observed, "When we bought our first tractor and retired our horses it was a big decision--but those who did not make that decision are not farming anymore". Has computerization been worth it2 Yes, we would do it again tomorrow! REFERENCES 1. ALLEN, R. R., COVEY.J. E. and M[NALDI,D. L. J. Am. Oil Chem. Soc. 51, 285A, Abstr. 99 (1974). 2. BECKWlTH, A. C., NIELSEN, H. C. and BUTTERFIELD, R. O. Analyt. Chem. 43, 1471-1474 (1971). 3. BIRKS, L. S. and GILERICH, J. V. Analyt. Chem. 48, 273R-280R (1976).

110

R. 0. Butterfield ef ul.

4 BITNER.E. D.. SNYI)ER,J. M.. ERNST. J. 0. and DUITON, H. J. Fefte Seijen Ansrr-Mittel 79, 483-486 (1977). 5 BRYANT,G. E.. SULLIVAN,F. and ROBE.C. Fd Process. 35, 5657 (1974). 6 BUTTERFIELD, R. 0. J. Am. Oil Chem. SOCK.46. 429431 (1969). 7 BUTTERFIELD. R. 0.. BITNER.E. D., SCHOLFIELD. C. R. and DUTTON. H. J. J. Am. Oil Chem. SOC. 41, 29-32 (I 964). x BUTTERFIELD. R. 0.. TJARKS,C. K. and DUTTON.H. J. Antdyt. Chem. 39, 497-500 (1967). 9 CRAM.S. P. and JUVET.R. S. Jr Anct/.rr. Chem. 48, 41 IR-442R (1976). IO CUSO. E.. GUARDINO.X.. RIERA.J. M. and GASSIOT. M. J. Chromatoyr. 95, 147-157 (1974). II DUFF. A. J. J. Am. Oil Chem. SM. 53, 370-381 (1976). 12 DUTTON.H. J. J. Am. Oil Chum. Sot. 51, 407-409 (1974). 13 DUTTON.H. J.. BUTTERFIELD. R. 0. and ROTHSTEIN, A. Anulyt. Chem. 38, 1773-1775 (1976). 14 Eight Peak Iude~ of Mass Spectra. Mass Spectrometry Data Centre. AWRE, Aldermaston, Reading, U.K. (1970). I5 ELLIOTT,P. J. Am. Oil Chem. Sot. 52, 127A. Abstr. 63 (1975). 16. ELLIOTT,P. J. Am. Oil Chem. Sot. 53, 15OA. Abstr. 127 (1976). 17. EMKEN.E. A.. ROHWEDDER, W. K.. DUTTON.H. J.. DOUGHERTY.R.. IACONO, J. M. and MACKIN. J. Lipids II, 135-142 (1976). IX FRAZER.J. W. Am. Lob. 5. 21-35 (1973). 19. GRAY, G. A. Amrf,vt. Chem. 47, 546A-564A (1975). 20. HELLER.S. R., MILNE, G. W. A. and FELDMANN.R. J. Science, N.E 195, 253-259 (1977). 21. HERTZ. H. S.. HITES,R. A. and BIEMAN.K. Anal.vt. Chem. 43, 681-691 (1971). 22. HITES.R. A. and BIEMAN.K. Analyt. Chem. 40, 1217-1221 (1968). 23. HITS. R. A. and BIEMAN.K. Analyr. Chem. 42, 855-860 (1970). 24. HOFREITER. B. T.. ERNST,J. 0.. ERNST.A. J. and RIST. C. E. TAPPI 51, 51A-56A (1968). 25. IBM Program 18GO-23.5.001. available from IBM, 40 Saw Mill Road, Hawthorne. New York 10532. R. and MCLAFFERTY,F. W. J. Am. &em. Sot. 95. 4185-4194 (1973). 26. KWOK. K. S.. VENKATARAGHAVAN, 27. LANSER.A. C.. MOUNTS,T. L. and EMKEN,E. A. Lipids 13, 103-109 (1978). 28. MCDONALD. R. S. Analyt. Chem. 48, 196R-216R (1976). 29. MCMANIS. G. E. and GAST, L. E. J. Paint Technol. 41, 581-582 (1969). 30. MOUNTS.T. L. and DUTTON.H. J. J. Am. Oil them. Sot. 44, 67-70 (1967). 31. PERONE,S. P.. ERNST,K. and FRAZER.J. W. Am. Lab. 5. 39-49 (1973). R.. DAYRINGER.H. E. and MCLAFFERTY,F. W. Anaiyt. Chem. 48, 32. PESYNA.G. M.. VENKATARAGHAVAN, 1362-1368 (1976). 33. Proceedings of World Conference on Oilseed and Vegetable Oil Processing Technology, .I. Am. Oil Chem. Sot. 53. 221462 (1976). 34. SELKE.E.. SCHOLFIELD. C. R.. EVANS.C. D. and DUTTON. H. J. J. Am. Oil Chem. SOC. 38, 614-615 (1961). 35. SNYDER,J. M., SCHOLFIELD. C. R.. MOUNTS, T. L., BUTTERFIELD. R. 0. and DUITON, H. J. J. Am. Oil Chem. Sot. 52. 244-247 (1975). 36. SWERN. D. (ed.) Bailey’s Industrial Oil and Fat Products, (3rd edn) Interscience. New York, 1964. 37. THOMAS,P. J. and DUI-TON.H. J. Analyt. Chem. 41. 657-660 (1969). 38. WARNER, K. A.. ERNST.J. 0.. B~UNDY. B. K. and EVANS,C. D. Fd Technol. 29(11), 42. 44-45, 47 (1974). 39. WASSON.J. R. and LORENZ.D. R. AnaIw. Chem. 48, 246R-261R (1976).