GINA-A Graphical Interactive Nuclear Magnetic Resonance Analysis Program

Stephen R. Heller

Heuristics Laboratory,

Division of Computer Research and Technology,

National Institutes of Health,

Bethesda Md. 20014

and

Arthur E. Jacobson

Laboratory of Chemistry, NIAMDD,

National Institutes of Health,

Bethesda, Md. 20014

THE AVAILABLE COMPUTER PROGRAMS (1--4) for calculating theoretical spectra are slow and cumbersome to use. Direct access to the WYLBUR text-editor system (5) on the DCRT/ Computer Center Branch IBM 360/370 batch programmed computer system at NIH enabled us to eliminate the error prone punched-card approach to the analysis program. However, in such systems the results from both the initial fit and iterative calculations are slow in forthcoming and the Calcomp plots of the results are even slower. While the computer run normally takes 1-2 hours, the overall elapsed time including the plots normally takes several days. The usual procedure for an initial fit calculation, with a number of first guesses, would take at least a week, or, on occasion, weeks, before a reasonable first fit is made for the experimental spectrum. In addition to the time factor, the cost of the numerous Calcomp plots and the calculations tend to make the analysis an expensive proposition.

Thus we decided to rewrite the original UEAITR program (4) for the PDP-10, a time-sharing computer. The advantages of the time-sharing computer are numerous, with the main ones being the possibility of direct and immediate interaction and a graphic display of the resulting calculation. We denote the composite of the original program, now capable of display and plotting of the initial fit and iterative calculations, as GINA (Graphical Interactive NMR Analysis).

Figure 1. Initial fit (zero iteration) of the data of Batterham, Bell, and Weiss at 250-Hz sweep width

Figure 2. The 100-MHz experimental spectrum of the ring C protons of codeine at 250-Hz sweep width

Figure 3. Initial fit (zero iteration) of the 100-MHz data at 100-Hz sweep width

The overall procedure [described in step-by-step fashion in an instruction manual (6)] for an analysis of an NMR spectral pattern is as follows:

1) Input trial parameters of chemical shifts and coupling constants using the PDP-10 text editor, SOS.

2) a) Run GINA for an initial fit.

b) Display results graphically.

c) Modify trial parameters, if necessary.

d) Run GINA for a new initial fit.

e) Etc.

3) Using the printed output of energy level labels and associated frequencies, text-edit the label and frequency disk file as needed, according to the frequencies noted in the experimental NMR spectrum. The user's manual (6) shows examples of the output of the initial run and the modified (text-edited) input for the iterative run. In this particular case, there were 80 theoretical line frequencies, 76 of which were used in the iterative calculation.

4) a) Run GINA for iterative fit.

b) Display results and, if satisfactory, plot the results of either the full spectrum, or portions of the spectrum as viewed on the display screen; the user changing the sweep offset and sweep width.

The total elapsed time for the above procedure is 1-3 hours. An example of the analysis of an ABCDX five-spin system in codeine will be given as an example of how the program works.

EXAMPLE

Table I. Chemical Shifts and Coupling Constants of the C Ring of Codeine

Figure 4. Final iterative fit of the 100MHz data at 250-Hz sweep width

The NMR spectrum of codeine (see Figure 2 for structure and numbering system) was desired in order to determine the configuration about C-6 of a newly synthesized 6-methylisocodeine and a spire oxirane from codeinone (7), and for the calculation of the shielding caused by the bonds in the Spiro oxirane molecule (8). Batterham, Bell, and Weiss have published an analysis of ring C of codeine using 60-MHz data (9). We used their values for the chemical shifts and coupling constants of the He' He, H., He, and Hl4 protons, and obtained the zero iteration (initial) fit shown in Figure 1. These were observed at the graphics terminal and plotted at a sweep width of 250 Hz. Although this spectrum is not overly different from the experimentally obtained spectrum shown in Figure 2, we desired a more accurate fit. Thus, we obtained the NMR spectrum of codeine on a 100-MHz instrument. The initial fit to the chemical shifts and coupling constants obtained from this spectrum is shown in Figure 3. This appeared to be somewhat better than the Batterham Bell, and Weiss fit, but was still not quite close enough to the Figure 2 spectrum. The printed output of frequencies obtained from this initial fit was modified to conform more precisely to the experimentally observed frequencies. After about three such trials, three iterations gave us our final result, now nicely matched spectra as shown in Figures 4 and 5. The Batterham, Bell, and Weiss parameters, along with the initial and final parameters found here are shown in Table I. All five chemical shifts and all ten coupling constants were varied in the iterative calculation, which required three cycles and 23 sec of cpu time to converge. The probable errors in the final calculated values in Table I are +0.020 to +0.031 and the RMS error is 0.106. The validity of using a different iso value for the H,4 proton (indicating a very different chemical shift from the other protons) was confirmed by the fact that in treating all the protons as an ABCDE system, the results (i.e., the energies, frequencies, chemical shifts, and coupling constants) were found to be the same.

It may be noted that Hl4 is not shown in the various figures, although its coupling constants and chemical shifts were used for the calculation. The codeine molecule, however, is rather complex; H,4 is further coupled to Hg, and the latter is coupled to the spins of the Hlo protons. Thus, H14 could not be obtained from these data, which were arbitrarily limited so as to achieve the restricted aims mentioned above.

Figure 5. Final iterative fit of the 100 MHz data at 100-Hz sweep width

COMPUTER PROGRAM

GINA is a modified version of the original UEAITR program. The program makes use of magnetic equivalence factoring to reduce the size of some matrices and, hence, increases the speed of the program. Iteration is done on the line frequencies as is done in the LAOCN3 (1) program, and the results can be output in a number of ways, including an

energy level diagram with lines assigned according to energy levels and/or connectivity.

UEAITR contained only a stick plot option. GINA contains the usual Calcomp plot option of the full spectrum drawn on a sheet of paper the exact size (50 cm) of the normal NMR spectrum chart paper. In addition, a plot option has been added that allows the user to plot that portion of the spectrum that is on the display screen. This latter plot is 25.9 cm by 25.9 cm. To get a display plot which can be directly overlapped with an experimentally obtained spectrum (Varian NMR instrument), the factor 1.93 (50.0 cm/25.88 cm) must be inserted for plotting purposes. The 100-Hz sweep width spectra reproduced in this paper were obtained via the graphic display plot, the 250-Hz sweep width spectral figures via Calcomp plots. While both contain identical information, and the latter are, perhaps, "prettier," these Calcomp plots require recomputing all the points on the graph, whereas the display plot uses the points already computed and displayed and is thus a faster plotting routine. The program, including the plotting, requires 48000 words of core to run. The display option, when used, requires 4000-8000 more words of core, depending on how much of the spectrum (how many points of the full curve) is being displayed on the screen. The entire program is written in FORTRAN IV and requests for further information should be addressed to S. R. H.

ACKNOWLEDGMENT

We would like to thank H. J. C. Yeh for the 100-MHz spectrum of codeine and J. A. Ferretti and M. McNeel for valuable discussions about the computer program.

RECEIVED for review May 2, 1972.Accepted July 31, 1972.

REFERENCES

(1) S. Casetellano and A. A. Bothner-By, J. Chem. Phys., 41, 3863 )1964)

(2) J. D. Swalen and C. A. Reilly, ibid., 42, 440 (1965).

(3) R. K. Harris and C. M Woodman, Mol. Phys., 10, 437 (1966).

(4) R. B. Johannesen, J. A. Eerretti, and R. K. Harris, J. Magic. Resonance, 3, 84 (1970).

(5) S. J. Kaufmann, A. E. Jacobson, and W. F. Raub, J. Chem. Doc., 10, 248 ( 1970); see references 2 and 3 therein.

(6) S. R. Heller, "DCRT/CIS, Gina--NMR Analysis Program Users Manual," Division of Computer Research and Technology, Bethesda, Md. 20014, May 1972.

(7) L. J. Sargent and A. E. Jacobson, J. Med Chem., 15, 843 (1972).

(8) A. E. Jacobson, H. J. C. Yeh, and L. J. Sargent, Org. Magn. Resonance, in press, 1972.

(9) T. J. Batterham, K. H. sell, and U. Weiss, Aust. J. Chem., 18, 1799 (1965).