independent of existing methods, additional water was added to two food products-instant coffee and vegetable oils. F o r instant coffee, 94% of the added moisture was quantitatively recovered while o n vegetable oils, 98% was recovered. Comparison with Other Moisture Methods. Table I1 shows the differences in moisture level determined by nearinfrared, standard five-hour-105 “C oven procedure, toluene distillation, and vacuum oven a t 60 “C and 27-inch Hg, o n a sample of instant coffee. The high values for both 105 “C oven and toluene distillation methods show the effects of thermal decomposition of carbohydrates. Under time and ternperature conditions, either method is capable of acceptable precision, but neither will give true moisture values. - Table 111 demonstratesthe ability of the near-infrared procedure t o be a n acceptable replacement for the Karl Fischer
method o n other food materials. A consistent high bias was noted on all moisture results of soybean oils determined by Karl Fischer as compared to near-infrared. The natural oil color presented interferences that made visual recognition of the true Karl Fischer end point difficult, and resulted in over-titration of the sample. As mentioned above, vegetable oil components such as peroxides and free fatty acids may also interfere in the Karl Fischer determinations. ACKNOWLEDGMENT
The authors thank C. C. Stophlet, W. L. Jasper, and R. W. Sanders for their suggestions and assistance in obtaining much of the analytical data. RECEIVED for review April 23,1970. Accepted July 10, 1970.
A Computer System for Use in Quantifying Liquid Scintillation Data Bruce E. Haissig and Arthur L. Schipper, Jr. North Central Forest Experiment Station, Forest Service-U. Folwell Auenue, S t . Paul, Minn. 55101
S. Department of Agriculture,
CONVERSION OF THE SEMIQUANTITATIVE DATA obtained by liquid scintillation spectrometry t o disintegrations per minute (DPM) or microcuries (pC) per sample entails laborious hand calculations. When two radionuclides such as 3Hand l4C are present in the same sample, and many samples are counted, the calculations become time consuming to the point of impracticality. Count data can be quantified, with varying degrees of automation, by means of programmable desk-top calculators (1-4) and on- ( 2 , 5 )and off-line ( 2 , 5 1 4 ) electronic computers. Although individual circumstances dictate use, the advantages of desk-top calculators and on-line computers (1) are generally outweighed by those of a n off-line computer and supplemental support services (2). Simply, the quantification and statistical analysis of large volumes of count data are most cheaply and accurately accomplished with a n off-line electronic com(1) M. F. Grower and E. D. Bransome, Jr., Anal. Bioclieni., 31,
159 (1969). (2) Y. Kobayashi and D. V. Maudsley. “Methods of Biochemical Analysis,” D. Glick, Ed., Vol. 17. Interscience Publishers, New York. 1969, p 55. (3) J. G. Manns and E. P. MacKenzie, Can. J . Physiol. P/iarmacol., 47,490 (1969). (4) B. F. Scott, J . Radioarid. Chem., 1,61 (1968). (5) J. H. Parmentier and F. E. L. Ten Haaf, Irit. J . Appl. Radiat. Isotopes, 20, 305 (1969). (6) J. M. Felts and P. A. Mayes, Biochem. J . , 105,735 (1967). (7) F. A. Blanchard, Int. J . Appl. Radiat. Isotopes, 14, 213 (1963). (8) M. I. Kiichevsky, S. A. Zaveler, and J. Bulkeley, Anal. Biochem., 22, 442 (1968). (9) C. Matthijssen, h d . , 15, 382 (1966). (10) R . Ninomiya, Ijit. J . Appl. Radiat. Isotopes, 17, 355 (1966). (11) J. J. O’Toole and J. 0. Oshurn, ibid., 19, 821 (1968). (12) E. D. Plotka, E. G. Stant, Jr., F. A. Waltz, V. A. Garwood, and R. E. Erb, ibid., 17,637 (1966). (13) J . L. Spratt, ibid., 16, 439 (1965). (14) J. L. Spratt and G. L. Lage, ibid., 18, 247 (1967). 1456
puter. This is especially true when a good interface is established between the scintillation spectrometer and computer (69). Several off-line computer programs that simplify the quantification of count data have appeared since 1963 (6-14). We were particularly interested in a program that would quantify, within established levels of error, both single- and doublelabel count data obtained from samples with a wide range of quench and varied radionuclide ratios. Anticipated low count rates dictated use of a program that would subtract variable levels of background, the result of differential quench among samples (15, 16). Simplicity in obtaining and using program “control” data was also desired because we have found that computer output is reliable only when control data are obtained anew for each experiment, even when counting parameters are standardized. The control data must protect the user against inherent instrument instability (which we have noted even over short periods), and against the changes in counting parameters resulting after recalibration of a n instrument that has been operated in a different mode. Finally, a program was needed that would accept data input as a single numerical expression of external standardization ( 2 , 5 , 17), such as dual-channels ratio (external standard ratio, ESR), and gross counts per minute (CPM) per sample. Available off-line computer programs did not meet all the above requirements regarding program capability, predictability of error, control data format, background subtraction, and mode of data input. We therefore developed a computer (15) I. T. Takahashi and F. A. Blanchard, Anal. Biochem., 29, 154 (1969). (16) D. L. Horrocks, “Survey of Progress in Chemistry,” A. F. Scott, Ed., Vol. 5. Academic Press, New York. 1969, p 185. (17) C. H. Wang, “The Current Status of Liquid Scintillation Counting,” E. D. Branson, Jr., Ed.. Grune and Stratton, New York, 1970, p 305 (in press).
ANALYTICAL CHEMISTRY, VOL. 42, NO. 12, OCTOBER 1970
program t o quantify digital output as gross C P M from liquid scintillation spectrometers that employ external standardization. Count data for single-labeled samples, double-labeled samples, or both can be quantified in a single run, even if methods of sample preparation or counting parameters are different. Gross C P M are corrected for variable background, counting efficiency, and where necessary, spillover (of the type defined below). The program is written in Fortran IV, with one version for the IBM 360 and another version for the CDC 6600. The general nature, use, and accuracy of the program are described here; the programs for both computers, along with specific instructions for use, are available o n request (please specify). EXPERIMENTAL
The following are required for single-label experiments : ESR and gross C P M for each experimental sample; ESR and counting efficiency (Eff) for each member in a series of increasingly quenched standards that contain the radionuclide present in the experimental samples; and ESR and C P M for each member in a series of similarly quenched background blanks. F o r double-labeled samples, if the radionuclide with the lesser maximum energy (A) is counted in channel 1 , and the other radionuclide (B) is counted in channel 2, the required data are: ESR, gross CPMA, and gross C P M B for the experimental samples; ESR, ERA, a n d Eff~for two series of quenched standards, each containing one of the radionuclides present in the experimental samples; ESR and background C P M in channels 1 and 2 for each member in a series of similarly quenched blanks; and ESR and spillover (S) for each of the quenched standards containing (B). S, for a given ESR, is the counting efficiency for B in channel 1. The method of using this control data is, within limits, at the discretion of the investigator. Equations for the quench correction, background, and spillover curves constructed from the control data (1) are not required by the program. Only some coordinates (up t o 99) for these curves are supplied t o the computer by means of control cards. These coordinates may be the unmodified original control data or, for example, coordinates may be generated from the original data by the method of least squares (1, 11). The range of ESR values entered o n the control cards should equal or exceed that of the experimental samples. It is also essential that the same scintillation medium be used in preparing standards and the experimental samples, and that the counting parameters be identical for each. Information on the program, control, and data cards (the latter in any numerical order) is relayed t o the computer, which, for a given ESR, determines the appropriate Eff, S, and background values by linear interpolation between coordinates supplied on the control cards. Then D P M and pC per sample are calculated by standard equations ( I ) . The computer printout shows D P M a , DPMB, pC of A, a n d pC of B, and the original data (ESR, C P M A , CPMB, a n d sample number). Several checks have been built into the program t o reduce error. The control cards are checked t o determine whether all necessary coordinates are present in the number specified o n the lead card. The data cards are also checked t o determine whether single- or double-label computations are required. If only single-label calculations are needed, the spillover routine is not used, so that data output remains free of meaningless numbers in the columns representing the unused channel. In addition, the data cards are checked for the presence of an ESR value. Data cards with missing ESR values are skipped and listed at the end of a run, without causing termination of subsequent calculations. Any number of data cards can be included in a deck. The insertion of new control cards between batches of data cards
signals the computer t o switch t o the new coordinates. This reduces computation costs, because re-entry of the program is avoided even though sample characteristics change, A test was made of the reliability of the computer program in estimating the true D P M originating from 3H a n d 14C in unquenched and quenched single- and double-labeled samples. A Beckman model LS-150 liquid scintillation spectrometer was used to obtain count data. The LS-150 has three samplecounting channels, and two additional channels for external standardization by means of the external standard counting ratio technique ( 2 ) using I3’Cs. Variations in the ESR ratio are related to changes in counting efficiency, such as those attributable to differential chemical quench, In addition t o external standardization, the LS-150 was equipped with Automatic Quench Control (AQC), which, under conditions of quenching, partially restores the energy spectrum of a radionuclide into a “window” with fixed discriminators (17). Spillover, though lessened, must still be accounted for in quantifying count data (1). Only two of the counting channels (1, narrow 3H Isoset; and 2, narrow l4C Isoset) were used to obtain count data. Isosets are counting windows with fixed discriminators. Use of the narrow windows minimized spillover of counts between windows, and this was further controlled as indicated below. The master gain potentiometer was adjusted so that about 0.25 % of the C P M originating from a n unquenched Beckman 3Hstandard, as determined with a wide-open window (the wide 14C Isoset), was recorded in channel 2 when the AQC was switched o n but its potentiometer set at zero. The ESR of the 3H standard was then dialed into the thumb-wheel resistors of the master gain control. Finally, the A Q C potentiometer was adjusted so that about 1.0% of the C P M originating from a highly quenched 3H standard (298,320 D P M 3H in 15 ml of scintillation medium containing 10% chloroform by volume) was recorded in channel 2. These adjustments are essential if the scintillation spectrometer is equipped for automatic gain restoration (17), such as AQC, but otherwise can be disregarded. The spectrometer, when adjusted as stated, counted standards quenched only with oxygen at efficiencies of 42 and 70% of 3H and l a c , respectively. Standards a n d test samples were prepared with a scintillation medium composed of butyl-PBD (8.0 g/L) and PBBO (0.5 g/L) in toluene containing known amounts of 3H-toluene, ‘4C-toluene, or both. The 3H standards (10 total), 14C standards (10 total), and background standards (5 total) were prepared with the same scintillation medium; and they contained these respective D P M : 298,320, 65,000, and 0. Chloroform was used as the quenching agent in amounts that yielded a fairly even distribution of ESR values between about 0.690 (unquenched) and 0.060 (most quenched) for each series of standards. Standards were counted t o a 2 u error of 0.2% or for 50 minutes, whichever occurred first. Only the original data, or efficiencies calculated from it, were used as control coordinates. Test samples were counted under the same conditions t o a n error of 1 % , or for 20 minutes. Count data were relayed from the scintillation spectrometer t o a model 33 Teletype which printed ESR and gross C P M in channels 1 and 2 for each sample, in addition t o punching the ASCII coded data into a paper tape. Computer punch cards were prepared with a n IBM model 47 tape-to-card printing punch. RESULTS AND DISCUSSION
I n general, the magnitude of error (resulting from sample preparation, counting, and program use) in the estimation of 3H or I4C in single-labeled samples, regardless of counting efficiency and level of radioactivity, was small and well within the allowable limits for most biological experiments (Table I). This, however, was not true for double-labeled samples, with which acceptably small errors were obtained only under re-
ANALYTICAL CHEMISTRY, VOL. 42, NO. 12, OCTOBER 1970
*
1457
Table I. Per Cent Error in Estimating DPM of 3Hand 14Cin Single- and Double-Labeled Samples0 Level of quench* Ratio of DPM Actual DPM Low Medium High 3H/'4C 3H 14 c 3H '4C 3H '4C 3H
... .., ... ... ...
*.. 0.0057 0.057 0.057 0.57 0.57 5.72 5.72 57.15
0 0 0 149 1,492 14,916 149 149 1,492 149 1,492 1,492 14,916 14,916
261 2,610 26,100 0 0 0 26,100 2,610 26,100 261 2,610 261 2,610 26 1
...
...
...
4.OC 0.9 -
1.9 81.2 14.1
6.5
1.3 0.7 0.5 2.2 1.5
5.0 6.3 5.0
... ... ...
6.1 6.5 6.7 5.0 5.9 9.6 6.9 21.8
6.1 5.9 5.7 *..
*.,
... ... 4.0
I .o 1.7 83.2 -
24.2 11.2 __ 0 2.4 0.3 1.7 1.0
...
..,
... 1.3
... ...
0.2 -
5.1 6.6 5.5 5.4 5.9 4.6 7.0 16.9
390.6 81.9 __
1.2
41.1 __ 13.4 7.4 2.6 0.5 0.5
c
_
'4C 7.0 8.1 7.5
... ... ...
7.0 8.0 7.2 6.1 7.7 9.2 8.9 25.3
Each value was calculated from mean estimated DPM of 5 samples. * Approximate respective counting efficiencies were 70, 62, and 51 for 14C; 41, 25, and 8 for 3H. Underline indicates per cent underestimate. a
stricted conditions. The amount of error encountered varied with aH/'4C ratio in a sample and its amount of quench. Small errors of estimate were obtained over the entire range of ratio of counting efficiencies tested only when the 3H/14C samples ranged from about 0.6 to 6.0. A tenfold increase or reduction in this ratio produced unacceptably large errors at all counting efficiencies. Very low ratios (0.0057 and 0.057) generally resulted in a n underestimation of 3H activity, while a very high ratio (about 57) yielded a consistent, rather large overestimation of 14Cactivity. Published reports of off-line electronic computer programs, at least those with which we are familiar, contain little error analysis, thus there is limited opportunity for comparing performance. Felts and Mayes (6) however, have reported relative errors encountered in the quantification of 3H and 14Cin serially quenched single- and double-labeled samples, the latter containing a nearly balanced radionuclide ratio. The errors we report are as small, over a greater range of quench. This was unanticipated, because the previous authors used a more elaborate spectrometer calibration procedure, coupled with a precise mathematical analysis of control data to produce required constants. There are two main sources of the observed systematic errors found in the estimates of the true activity of 3H and '4c in double-labeled samples. At low 3H/14Cratios, even small errors in the determination of 14C D P M cause pronounced inaccuracy of the spillover term in the equation, which in this and other programs reduces the inflated 3H C P M before corrections are made for background and counting efficiency. Overestimation of 14CD P M , such as we encountered, results in a n underestimation of 3HD P M . As the level of quench increases, it is more difficult to properly estimate 14CD P M , and the 3HC P M also increases, so that the quantification of 3H count data becomes increasingly less accurate. Conversely, errors in the estimation of 14Cactivity result when the 3H/14C ratio is very high because the calculations used in the computer program d o not include a correction for spillover of 3H C P M into the channel. Since the scintillation spectrometer was standardized t o maintain this spillover as a constant percentage of the total 3H CPM, low 14Cactivity is erroneously increased by readily measured amounts when 3H activity is relatively much larger. 1458
Accuracy of the program is greatly dependent, other conditions being equal, o n the precision of values derived through linear interpolation by the computer from the quench correction, background, and spillover control data. I n the program test reported here, n o attempt was made to maximize the precision of these derived values, because we only wished to illustrate average results that can be obtained over a wide range of quench and radionuclide ratios. In actual practice, particularly when double-label data are quantified, the user will want to maximize precision of the control data as follows: 1. Where instrument design and time allow, adjust the spectrometer such that linear or near linear quench correction curves are obtained for each radionuclide. 2. Use as many standards as deemed feasible t o obtain control coordinates, and restrict the quench of standards to the range found in the experimental samples in order t o increase the number of functional coordinates. 3. Count several standards at each level of quench, and use mean ESR and Eff values as coordinates. This also allows the establishment of confidence limits and aids in step 4. 4. Generate secondary control coordinates from quench correction, background, and spillover curves fitted by the method of least squares.
We did not include a statistical treatment of quantified count data in the program because greater flexibility is obtained when such analyses are performed by the readily available programs in most computer libraries. An interface between the data quantification and statistical programs is conveniently established by requesting punch card data output, followed by any required machine sorting of the cards. ACKNOWLEDGMENT
We thank Mr. David Schempp, University of Minnesota, for writing the program for the IBM 360; Dr. James P. King of this Station for aiding in development of this program; and Mrs. Ellen M. Peterson of this Station for adapting the program to the CDC 6500.
RECEIVED for review May 7, 1970. Accepted July 15, 1970.
ANALYTICAL CHEMISTRY, VOL. 42, NO. 12, OCTOBER 1970
