mass spec 1

8/3/2019 mass spec 1

1/7

1702 Anal. Chem. 1989, 61, 1702-1708

Interpreting Mass Spectra of Multiply Charged IonsMatthias Mann, Chin Kai Meng, and Jo hn B.Fenn*Department of Chemical Engineering, Yale U niver sity, New Haven , Connecticut 06520-2159

We descrlbe two algorHhms that extract molecular mass In-formatlon from spectra showlng sequences of peaks due toions wRh varylng numbers of charges. The flrst, called herethe averaglng atgortthm, unambiguously asslgns chargenumbers to the Ions assoclated with the m / z value for eachpeak In the sequence and then averages the resuttlng valuesof M to glve a best eotlmate d the molecular mass. Thesecond, ldentlfled as the deconvolutlon algorlthm, mathe-matically transforms a spectrum of several peaks for muttlplycharged lono Into one peak correspondkrg to a slngiy chargedion. The procedures can be readily Implemented wRh apersonal computer and are here applied to representatlvespectra of small proteins generated by electrospray massspectrometry. These algorl#rms are now routinely used In ourlaboratory for the lnterpretatlon of such spectra. They bothare fast and convenlent,discriminate agalnst beckground, andtake advantage of much of the lnformatlon contalned In asequence of peaks. Achievable accuracy and sources oferror are discussed.

I. I N T R O D U C T I O NThe ions produced by the sources traditionally used in massspectrometry generally comprise singly charged species re-sulting from the loss or gain of an electron by a parentmolecule. Moreover, an appreciable fraction of the ions areoften charged fragments of the parent molecule. On the otherhand, ions produced by some of the more recently developedsources consist of neutral parent molecules to which smallcations or anions are attached. Among these newer andsofter ionization methods are electrohydrodynamic ioniza-tion (EH), fast atom bombardment (FAB), fast ion bom-bardment (FIB) commonly referred to as secondary ion massspectrometry (SIMS), laser desorption (LD), plasma desorp-tion (PD), thermospray (TS), and aerospray (AS) originallyknown as atmospheric pressure ion evaporation (APIE). Duein part to the larger size of the molecules that can be accom-modated by these new sources and in part to the nature oftheir ionization processes, ions containing up to five or sixadduct charges have been observed ( 1 ) . However, to ourknowledge, except for some preliminary work in our labora-tory, no study on how to make efficient use of the peakmultiplicity has been reported (2).Recently, with an electrospray (ES) mass spectrometer thathas been previously described (3),we have been able to obtainthe mass spectra shown in Figure 1 for eight small proteinswith molecular weights from 5000 to almost 40000 ( 4 , 5).Analyte samples were dissolved in solvents comprising mix-tures of acetonitrile, water, and methanol or 1-propanol. I twas necessary to lower the solution pH by addition of smallquantities of acetic acid (HAC)or trifluoroacetic acid (TFA).The optimum proportions of these solvent components de-pended somewhat on the particular sample and were deter-mined by trial and error. Solutions with analyte concentra-tions ranging from 0.7 to 13 7 pmol/L, depending upon thespecies, were injected at flow rates of 8 pL/min. Each of thespectra shown is the result of a single scan requiring 30 s tocover the indicated mass range. The analyzer was a VG

Micromass 1212 with a nominal upper limit for m / z for 1500.The analog output from the Channeltron detector was digi-tized with an analog to digital converter and fed into ahomemade data recording and processing system based on anIBM-AT clone. Since our preliminary report at t he ASMSMeeting in San Francisco last June , two other groups haveconfirmed our results ( 6 , 7 ) . Indeed, Edmonds et al. were ableto obtain ES spectra for a bovine albumin dimer with a mo-lecular weight of 133000.Although the experiment was not optimized for sensitivity,it is apparent from Table I and Figure 1 that very low de-tection limits can be achieved. For example, the spectrumof lysozyme consumed only about 3 pmol of sample althoughmore was used because processing and manipulation were notvery efficient. In each case the spectrum comprises a sequenceof peaks with an intensity distribution that isnear Gaussian,has a width of around 500 on the m / z scale, and is generallycentered at a value between800 and 1200. The constituentions of each peak differ from those of it s adjacent neighborsby one elementary charge. For the readers convenience wehave shown the number of such charges per ion for two orthree peaks in each spectrum. Each such charge is due to anadduct cation from the original solution. Our analyzer didnot have sufficient resolution for large ions at these m /z valuesto permit an unequivocal assertion of unit mass for an adduction. However, the need for low pH in the sample solution,along with results obtained for smaller peptides and aminoacids, strongly support our assumption t ha t H+ s the mostlikely charge carrier in these experiments.For the eight proteins we studied Table I summarizes theessential features of each spectrum and the information itprovides. It is immediately apparent from the figures and thetable that the degree of multiple charging in ES ionizationismuch higher than has been encountered with any other softionization method. This feature is very attractive in that itextends the effective mass range of any analyzer by a factorequal to the number of charges per ion. -Moreover,becausethe ions have lower m / z values, they are generally easier todetect and weigh than are singly charged ions of the samemass. On the other hand, peak multiplicity distributes thesignal for one species over several channels. But because thenumber of charges per ion is almost always greater than thenumber of peaks, the tota l current carried by one species isgreater when there is peak multiplicity than would be the casefor a single peak containing the same total number of singlycharged ions. Unfortunately, we do not yet know the detectorresponse per charge of a multiply charged ion. We do know,however, that no postacceleration has been required formultiply charged ions tha t were large enough to require suchacceleration had they been singly charged. We also know thatthe detection sensitivity obtained with ES ionization of largemolecules seems to be substantially greater than has beenobtained with sources giving rise to ions tha t are predomi-nantly singly charged (8). Moreover, as will emerge in thesubsequent discussion, because peak multiplicity allows signalaveraging, mass assignment can be made with more precisionand confidence than would be the case for a single peak ofa singly charged ion. The objective of this paper is to presentbasic methods for interpreting the sequence of multiplycharged peaks and to provide algorithms for retrieving the

0003-2700/89/0361-1702$01.50/00 1989 American Chemical Society


2/7

ANALYTICAL CHEMISTRY, VOL. 61, NO. 15, AUGUST 1, 1989 1703Table I. Data for the Spectra in Figure 1

concentrationmol w t g/L Fmol/L charges m / z range

insulin (bovine) 5 733 0.05 8.8 4- 6 950-1450cytochrome c (horse heart) 12 360 1.67 137 12-20 600-1100lysozyme (chicken egg) 14 306 0.01 0.71 10-15 900->1500myoglobin (equine skeletal muscle) 16 950 1.00 58.8 15-27 600 -1400trypsin inhibitor (soybean) 2009 1 0.10 5.0 16-22 800-1400a-chymotrypsinogenA (bovine pancreas) 25 656 0.50 19.0 17-22 1150-> 150 0alcohol dehydrogenase (horse liver) 39 830 0.50 12.5 32-46 800-1300carbonic anhydrase I1 (human erythrocytes) 29 006 0.50 17.2 23-36 725-1500

The molecular weight was determined from the sequence information provided mostly by ref 8 and is an average value based on thenatural abundance of isotopes.

0 400 800 1200 1600

'11 Lysozyme M.W. = 14,30664 -2 -0

- 15"1 1 , 1 1 , 10 400 800 1200 1600

8l o Typrln lnhlbltor

M.W. = 20.091

l o 1 Carbonlc Anhydrare II8 M.W. = 29,0066420700 900 1100 1300

500 700 900 1100 1300

where q is an elementary charge and i is unity in conventionalspectra for singly charged ions. It should be kept in mind thatthe units of m/ z are properly daltons per elementary chargeeven though a measured peak position is often loosely ex-pressed simply in daltons (Da) when z is one. All the formulasapply equally well to negatively charged ions with mabeingnegative in the case of charge abstraction. Thus one can writefor each of the peaksI O

86420400 800 1200

11 a-Chyrnolyprlnogen AM.W. = 25,656iu00 1000 1200 1400 1600

Alcohol Dnhydrogenase~ 5 H + M.W. = 39.8306420700 900 1100 1300 1500

Figure 1. Spectra of multiply charged ions produced with electrasprayionization (ES). Each of the spectra was acquired with a single, 30-9scan. The number of charges ia re indicated for some representativepeaks. Table I ghres some data on these spectra. See ref 3 for moredetails on the experiments.mass information they contain. Two such algorithms, alongwith illustrative results obtained by applying them, will beset forth in the following paragraphs. In all the calculationswe have assumed that the detector response to any ion didnot depend on the number of its charges.

11. A V E R A G I N G A L G O R I T H MIf one assumes that in a particular spectrum the adductcharges of each ion all have the same identity, and thereforemass, and th at any neutral adducts such as solvation speciesare the same for each ion, then there are three variablesassociated with each of the peaks in the series: the mass M(numerically equal to the molecular weightM,)f the parentmolecule including neutral adducts, the number of chargesi, and the mass maof the adduct ions. We use i rather thanz to designate the number of charges in order to avoid con-fusion with the customary m / z scale of mass spectra for whichm = M + im,, the total mass of the ion. In general,z = i q ,

or

where Ki is the apparent value of m / z for the peak positionon the scale of the mass analyzer and K 0) yields for the number ofcharges i

For example, if the adduct ions are protons (m a= 1) andwe observe one peak a t Ki = 1001.0 and another one two peaksaway (j = 2) at Ki+,= 834.3, then we would get i = 2K


3/7

1704 ANALYTICAL CHEMISTRY, VOL. 61, NO. 15, AUGUST 1, 1989and further justified below, an assumption tha t ma = 1 seemsto be appropriate (at least as long as there is not too muchsodium in the solution) and will be made in the numericalexamples in this paper.With known charge number, and measured or assumedadduct ion mass ma, the parent ion mass M can be obtainedfrom any one peak or averaged from a number of peaks

1M = --CiK:no i (4 )where the summation is over the i values for the peaks selededfor averaging and no s the number of those peaks.The coherence of the peak sequence makes possible furtherimprovement in the estimate of M given by eq 4. I t allowsus to identify and ignore peaks that do not belong in thesequence and to evaluate the quality of the spectrum. Fromeq 1 for any two peaks we obtain

K : i- l + :-Ki+j 1Hence any pair of peaks in an experimental spectrum de-

fines a point withy = [ ( K : / K : + , ) l ) / j nd x = l l i . All suchpoints should fall on the line y = x . The scatter of thepair-points around this line is a measure of the quality of thespectrum. If they were all precisely correct, they would allfall precisely on the line. Figure 2a shows such a plot for thecytochromec spectrum of Figure 1. The seven points a t eachabscissa value of lli correspond to the seven possible ratiosof KJK


4/7

ANALYTICAL CHEMISTRY, VOL. 61, NO. 15 , AUGUST 1, 1989 1705- 1.5

- 1.0' 5

- 0. 5

- 8 -6 - 4 - 2 0 2 4 6 8A m (D)

Flgure 3. Results of offsetting m l r for the i = 19 peak by an amountAm n the spectrum for cytochrome c. The dashed line shows theeffect on unweighted average mass, the solid line on the weightedaverage. Note that the weighted average is much less affected by theoffset peak once that peak is outside its "best position" with respectto the rest of the sequence. The open circles represent (on theright-hand ordinate) the relative weighting factor w , ~hen p = 2.Ta ble 11. Comparison of Unweighted and WeightedAverages of Experimen tal Values for Molecular Weight ofCarbonic Anhydrase Whose True (Sequence) Mass i s 29006weighting index p

02 28984.2 12.23 28985.6 f 10.94 28987.3 9.85 28989.1 f 8.86 28990.8 * 7.9

M, , (* td dev)28 982.2 f 15.6 (unweighted)

In comparing the parent mass obtained by this weightingprocedure with the true mass, one has to keep in mind tha tthere are at least two sources of error tha t contribute to AM(i.e. Mme - M m ~ ) .ne, the statistical error in ascertainingthe individual peak positions is expressed in the unweightedor weighted standard deviation of the measured massM. heother arises from systematic errors in the calibration of theanalyzer mass scale. This latter source of error will obviouslynot be affected by any weighting procedure. If the error dueto mass calibration predominates, weighting the average willnot provide a major improvement in mass accuracy. In sucha case the standard deviations, weighted or unweighted, donot indicate the experimental accuracy of a measurement butonly its precision. A criterion for deciding if mass scalecalibration is negligible in determining the error in M is

(i,) ADa


5/7

1706 ANALYTICAL CHEMISTRY, VOL. 61, NO. 15, AUGUST 1, 19891 " ' " " " " " ' " " " ' I

M / 6 i0.00 0.05 0.10 0.15 0.20

m/z/ M

L M

0.0 0.5 1.0 1.5 2.0M yFigure 4. (a)A synthetic sequenceof peaks or multiply charged ionswith 6 to 15charges. (b) Deconvolution of (a)using eq 8. The massscale is in units of the parent mass M .simplified in the coherent sequence of peaks could betransformed to one singly charged peak located on an m/zscale at the molecular mass M of the parent compound. Itwill be shown that the following function can provide sucha transformation:

F is the transformation function for which the argument M*is any arbitrarily chosen trial value of M for which F is to beevaluated. The symbolf represents the distribution functionfor peak heights in a measured spectrum. For example, ifthere is a peak of relative intensity 5 at m/z = 500 then f (500)= 5. ma is the adduct ion mass, as previously defined. It willbe shown below tha t the function F has its maximum valuewhen M* = the actual value of M, he parent mass of the ionsof the peaks in the sequence. Thus, evaluating F a t all valuesof M* with 0 5 M*I yields a transformed or "deconvoluted"spectrum, in which the peak with maximum height corre-sponds to the parent species with a single massless charge.An example will make it clear how the deconvolution algor-ithm reconstructs the parent peak from the sequence. Forsimplicity we assume ma = 0. Figure 4a shows a hypotheticalmeasured spectrum f generated by charging a molecule withmass M with from 5 to 15 massless adduct ions such that theheight is unity for every peak in the sequence. These"measured" peaks occur at M/6 , M/7 , ..., M/15. If F isevaluated a t M* = M the following sum is obtained:F ( M ) =4:) + f f ) ... + f( f) +4:) + ... +

f( g ) + f( E ) + fg) * e .= o + o + ...+ 0 + 1 + . . .+ 1 + 0 + 0 + ...= 10

Thus, the function F has created a peak a t the position M*= M with a height equal to the sum of the heights of thesequence peaks. As noted earlier we do not yet know howdetector response depends on the number of charges per ion.In all the work discussed here we have assumed that the heightof any peak in a measured spectrum is related to the abun-

dance of it s ions by the same proportionality constant nomatter how many charges are on those ions.If F is evaluated at M + t , a position slightly larger thanM, then F will be zero because (M + t) does not correspondto the position of any of the sequence peaks. However, it isalso apparent from the example that the functionF will createpeaks in the deconvoluted spectrum a t more positions thanat M* = M. A t M* = ( 2 / 3 ) M the following sum results:

Figure 4b shows the results of applyingeq 8 to the spectrumof Figure 4a, an ideal sequence of multiply charged ions with6I I 15. It is a property of the spectrum resulting fromthe transformationF, as shown in Figure 4b, that it comprisesa series of calculated peaks containing contributions from theactual peaks in an observed spectrum. By reference to theabove procedure, it is easy to infer a number of general fea-tures of the deconvoluted spectrum. As we have already noted,its most prominent peak occurs when M* equals the parentmass M and has a magnitude equal to the sum of the mag-nitudes of the individual peaks in the sequence. The nexthighest peak occurs at M / 2 and it is a t most only half aa highas he peak at M. In general there are peaks at (k/ i)M,wherei,, 5 I i and k is any integer. In the sequence of "sidepeaks" on either side of the parent peak those closest to theparent (maximum) peak M occur at ((i- f )/i-)M wherei,, is the highest number of charges on a single ion. Theposition of these closest side peaks is indicated by arrowheadsin Figure 4b. The height of these side peaks is a factor ofl / ( i , - ~ Jmaller than the height of the molecular peakat mass M. It also turns out that the deconvoluted spectrumis periodic in M. This periodicity may be viewed physicallyas being due to synthet ic "overtones" of the basic spectrumcorresponding to doubling, tripling, etc. of both the parentmass and the number of charges on each peak, and a differenceof 2, 3 , etc. in the i values of adjacent peaks.The transformed spectrum changes somewhat in appear-ance if finite resolution and background are taken into ac-count. To simulate these effects, the shapes of individualpeaks in the sequence of Figure 4a are represented in Figure5a by isosceles triangles with a relative fullwidth at half height(fwhh)of 0.005. Furthermore, a constant background of 10%of the peak height was introduced. The consequences of thistreatment are seen in Figure 5b. There is a progressive in-crease in the magnitude of the "side peaks" because thenonzero peak width in the observed spectrum results in acontribution to F at m / z values on either side of the peakcenters. The steady increase in the "base line" is caused bymore frequent sampling of the background a t higher valuesof M *.This "deconvolution algorithm" has been programmed forboth an IBM-AT clone and a Macintosh SE. It was appliedto the eight experimental spectra shown in Figure 1with theassumption tha t ma = 1. Computation time was usually lessthan 1 min. Each mass spectrum was represented by 1150points for a full scan. In the algorithm a linear interpolationbetween adjacent data points was used. It should be pointedout that this algorithm needs no a priori information aboutcharge states or the number of peaks in the sequence. Theonly instruction specific to a particular spectrum is the rangeof m / z in the "window" that spans the peaks to be deconvo-luted. Thus the summation of eq 8 goes only from the mi-mimum to the maximum values of m/z within this window.


6/7

ANALYTICAL CH EMISTRY, VOL. 61, NO. 15, AUGUST 1, 1969 0 1707Table IV. Summary of Values for Molecular Weight (M)rom the Spectra in Figure la

Md (byM , (unweighted av) M* (weighted av, p = 2) deconvolution)species M , M , i td dev AMr, % MW,(2) A .Mru , ( Z ) r 9% Md md,%

insulincytochrome clysozymemyoglobintrypsin inhibitorba-chymotrypsinogenAccarbonic anhydrase I1alcohol dehydrogenase

5 73312 36014 30616 95020 09125 65629 00639 830

5734 i 1412349 i 514324 f 1516906 i 1119990 f 3726131 f 228982 f 1639859 f 25

0.010.090.130.260.501.80.080.07

5740 f 1212350 f 414329 f 1016904 i20001 f 2826 130 i 228984 f 1239871 f 17

0.12 5 7510.08 12 3520.16 14 3400.27 16 9270.45 20 0231.8 25 9390.08 29 0050.10 39 876

0.310.070.240.260.341.10.0040.12

"All M values are isotope-averaged. M . is from sequence (ref lo),Mr is unweighted av. Ma s weighted av (p = 2). *75%of the moleculesare said to lack the terminal Leu (ref 10). If true in our sample,AM's would be much smaller. CThismay be an atypical case since there areonly four broad peaks to average and the mass window extends beyond m / z = 1500 (see Figure 1). However, the standard deviationindicates that the measurement error in this spectrum should be only slightly higher than in the other spectra.

0.00 0 . 0 5 0.10 0.15 0 . 2 0ml z (D)

2 MI

Figure 5. (a) Synthetic sequence of peaks whose shapes are ap-poxknated by isosceles triangles (fwhh = 0.5 %). There is a constantbackground that Is 10 % of the peak height. (b) Transformation of (a)using eq 6 . The mass scale is in units of the parent mass M.Such a limitation in the range of the summation reduces thenoise in the transformed spectrum because background thatlies outside the range of interest is not sampled. Figure 6adisplays the result of applying the deconvolution procedureto the spectrum of cytochrome c inFigure 1. The transformedspectrum clearly shows the side peaks, the overtone period-icity, and the base line increase discussed above. The parent(largest) peak is magnified in Figure 6b by "zoom" expansionof the mass scale in its vicinity. Figure 7 shows the resultsof the same treatment for the case of a larger protein, carbonicanhydrase I1 (M= 29 006). Widths at half maximum for bothmeasured and deconvoluted peaks for the other spectra wereusually about 1% . Such large spreads resulted in part becauseeffective resolution of our analyzer was only about 300. TableIV summarizes the results for molecular weight determinationfrom the spectra in Figure 1.It is interesting to note that in general there is a regionimmediately around the parent peak that is free from artifactsof the deconvolution algorithm. As noted above, this regionshould extend from (imm- ) / im mX M to (imm+ l ) / i m mXM, where i is the maximum number of charges found ona molecule. In Figures 4 and 5 the boundaries of this regionare marked by black triangles. In the deconvoluted cyto-

1 '"1d 4 0

v i I5 0 0 0 1 0 0 0 0 1 5 0 0 0 2 0 0 0 0 2 5 0 0 0 3 0 0 0 0

Mn (D)Figure 6. (a) Deconvolution of the cytochrome c (M = 12360)spectrum of Figure 1 using eq 8 with ma= 1. The theoreticalpositionsof the first side peaks are marked by dark triangles. (b) "Zoom"expansion of the spectrum in (a) for the mass range between 10 000and 14 000. For explanation of the peak marked by the open trianglesee text.chrome c spectrum (Figure 6), however, a small peak (markedwith an open triangle) is observed about 340mass units higherthan the molecular peak. Detailed examination of the mea-sured spectrum reveals a small peak just above each main peakin the sequence whose position agrees with the peak foundin the deconvolution. This observation indicates that thealgorithm can readily detect small peaks closeto a parent peakthat may be due, for example, to parent species variants withslightly different masses.The question arises as o the complications introduced whentwo or more parent species are present in the sample. In somepreliminary experiments, with mixtures having two compo-nents, e.g. cytochrome c and myoglobin, we have been ableto resolve their spectra sufficiently to obtain mass values with


7/7

1708 ANALYTICAL CHEMISTRY, VOL. 61, NO. 15 , AUGUST 1, 1989

20000 30000 40000 50000M* (D)

6 0 MA I

2 6 0 0 0 2 8 0 0 0 3 0 0 0 0 32000 34000M* (D)

Flgwe7. Results as in Fgure 6 of applying the deconvdutionalgorithmto th e carbonic anhydrase I I (M = 29 006) spectrum of Figure 1.the same accuracy and confidence as if each componentpresent had been separately analyzed. We plan further ex-periments with mixtures and will report on them later.

IV . C O N C L U S I O NTwo procedures have been presented for interpreting

spectra comprising sequences of peaks for multiply chargedions. One of them establishes unambiguously the number ofcharges for each measured peak and discriminates againstbackground peaks that do not belong in the sequence. Therelevant peaks can then be averaged with confidence to cal-culate the molecular weight. The other algorithm transformsa measured spectrum and yields directly the peak tha t wouldbe expected for a parent species with a single massless charge.The methods complement each other. The averaging algor-ithm will be most useful in dealing with a noisy spectrumhaving relatively sharp peaks. It will also be the method ofchoice when a sequence contains only a few large peaks and

many small ones because each peak can then be given the sameweight in the averaging. A spectrum transformed with thedeconvolution algorithm, on the other hand, in effect weightsthe sequence peaks by their relative magnitude and thus willbe most responsive to those with large amplitudes. Thatalgorithm retains some peak shape information and can alsoresolve mixtures. Often it may be appropriate to use bothand to compare their results. For example, if one is notabsolutely sure whether a particular peak M in the decon-voluted spectrum is the true parent, one could use the aver-aging algorithm to decide. Both algorithms discr iminate ef-ficiently against background and use much of the informationavailable in the measured spectrum. As we have demonstratedhere, ES mass spectrometry combined with each of theseprocedures generally allows rapid and confident determinationof molecular weights for large molecules with an accuracy ofa few tenths of a percent or better, even with instruments ofmodest resolution. With suitable analyzers this accuracy couldbe significantly increased.

A C K N O W L E D G M E N TSpecial thanks are due to Shek Fu Wong, Craig M. Whi-tehouse, Michael Labowsky, Peter Chen, and Walter J.McMurray for many stimulating and fruitful discussions. Wealso register our appreciation for advice, encouragement, and

samples from other colleges at Yale too numerous for a rollcall.L I T E R A T U R E C I T E D

(1) Roepstorff,P.; Sundquist, B. In Mass Spectrometry in Biomedical Re-search; Gaskell, S. J., E d.; John Wlley 8 Sons Ltd.: London, 1986; pp269.(2) Mann, M.; Meng, C. K.; Fenn, J. B. Proceedings of the 36th AnnualConference on Mass Spectrometry and Allied Topics 1988, 1207.( 3 ) Whitehouse, C. M.; Dreyer, R. N.; Yamashka, M.; Fenn, J. B. Anal.Chem. 1985, 57 , 675.(4) Meng, C. K.; Mann, M.; Fenn, J. 8. Proceedings of the 36th AnnualConference on Mass Spectrometry and Allied Topics 1988, 771.( 5 ) Meng. C. K.; Mann, M.; Fenn, J. B. 2. hys. 1988, 70, 361.(6) Covey, T. R.; Bonner, R. F.; Shushan, B. I. ; Henion, J. Rapid Com-mun. Mass Spectrom. 1988. 2 , 249.(7 ) Edmonds, C. G.; Loo, J. A.; Barlnaga, C. J.; Ws eth, H. R. ; Smlth, R. D.4th International Symposium on LC-MS, Freiburg, November 1988.(8) Barber, M.; Green,B. N. Rapid Commun. Mass Spectrom, 1987, 7 ,80 .(9) Yergey, J. ; Heller, D.; Hansen, G.; Cotter, R. J.; Fenselau, C. Anal.Chem. 1983, 55 , 353.(10) Atlas of Protein Sequence and Structure; Dayhoff, M. O ., Ed.; NationalBiomedical Researc h Foundation: Washington, DC, 1972-78; Vol. 5and Suppl. 1-3.

RECEIVEDor review August 15,1989. Revised manuscriptreceived January 24 , 1989. Accepted April 24, 1989. Thisresearch was supported by the National Institutes of Healthunder Grant 5 R 0 1 Gm31660-3. Partial support was alsoprovided by the donors of the Petroleum Research Fund,administered by the American Chemical Society.

mass spec 1

Documents