Identification of volatile and semivolatile compounds in chemical ionization GC-MS using a Mass-To-Structure (MTS) Search Engine with integral isotope pattern ranking

Abstract

The mass-to-structure or MTS Search Engine is an Access 2010 database containing theoretical molecular mass information for 19 438 compounds assembled from common sources such as the Merck Index, pesticide and pharmaceutical compilations, and chemical catalogues. This database, which contains no experimental mass spectral data, was developed as an aid to identification of compounds in atmospheric pressure ionization (API)-LC-MS. This paper describes a powerful upgrade to this database, a fully integrated utility for filtering or ranking candidates based on isotope ratios and patterns. The new MTS Search Engine is applied here to the identification of volatile and semivolatile compounds including pesticides, nitrosoamines and other pollutants. Methane and isobutane chemical ionization (CI) GC-MS spectra were obtained from unit mass resolution mass spectrometers to determine MH⁺ masses and isotope ratios. Isotopes were measured accurately with errors of <4% and <6%, respectively, for A + 1 and A + 2 peaks. Deconvolution of interfering isotope clusters (e.g., M⁺ and [M − H]⁺) was required for accurate determination of the A + 1 isotope in halogenated compounds. Integrating the isotope data greatly improved the speed and accuracy of the database identifications. The database accurately identified unknowns from isobutane CI spectra in 100% of cases where as many as 40 candidates satisfied the mass tolerance. The paper describes the development and basic operation of the new MTS Search Engine and details performance testing with over 50 model compounds.

---

Introduction

There has been considerable interest in “spectrumless” databases to assist in identification of unknown samples in API-LC-MS.^1–3 These databases contain the theoretical monoisotopic masses of API-LC-MS ions such as MH⁺ and [M − H]⁻. Spectrumless databases have become popular because of the availability of high mass accuracy instruments such as time-of-flight (TOF) and Orbitrap instruments. Moreover, because they do not contain experimental data, they do not require authentic analytical standards which may be expensive or difficult or impossible to obtain. Unequivocal identification still requires analysis of authentic standards, however.⁴

At this time there is no standardized workflow, but a popular variation has been termed “suspects screening”⁵ where multiple extracted ion chromatograms mine MS datafiles for analytes of interest. Narrow mass ranges (e.g., ±10 ppm)⁶ provide adequate specificity, at least for purposes of screening. The extraction mass window depends on the instrument mass resolving power.⁷ Suspects screening has found numerous applications in forensics including determination of street drugs,⁸ androgens in biological specimens,² and doping agents in urine.⁹ Suspects screening has also been used in environmental applications such as the determination of unknown pesticides and degradates in food^10,11 and over 1500 microbial transformation products of surface water contaminants.¹² Additionally, similar techniques have been used for metabolomic studies of human urine using a 15 000 compound metabolite library.¹³

An alternative approach to suspects screening is “nontarget screening”⁵ in which there is no a priori information on analytes and no list of masses to search. A typical workflow in nontarget screening includes: automated peak detection; creation of candidate empirical formulas; query of a database(s) such as Pubchem to obtain candidate structures; and further ranking by MS-MS fragmentation.⁵ Empirical formula discovery is not required by all spectrumless databases, however.¹ Suspects screening is an example of a reverse library search, while nontarget screening and the MTS Search Engine are forward search methods.

The present study describes further development of the MTS Search Engine. Here the database was modified to incorporate automated consideration of isotope patterns, providing needed orthogonal information.^5,6,12 Secondly, the MTS Search Engine was adapted for identification of volatile and semivolatile samples using chemical ionization GC-MS. This is a novel extension of the applicability of spectrumless databases beyond API mass spectrometry of nonvolatile analytes. Because the search engine relies on molecular mass information, CI was preferred over EI mass spectrometry. In this study the effectiveness of the new database was judged using data from single quadrupole instruments that afford only unit mass resolution.

At this time the identification of volatile compounds relies almost exclusively on computerized spectral matching using 70 eV, electron ionization (EI) spectral libraries.¹⁴ Identification is based upon a fingerprint of mass spectral ions and fragments, and this approach is effective because of the reproducibility of EI mass spectra and the availability of large EI spectral libraries and reliable algorithms for assessing spectra similarity and dissimilarity. While both powerful and easy to use, searching experimental MS libraries has limitations that can lead to false identifications.^15–17

Identification errors stem from a number of factors. The available spectral libraries contain a few hundred thousand records including the majority of environmentally important compounds, but they do not contain the millions of compounds possible. Multiple compounds may be present in an unknown mass spectrum due to the coelution of two or more components. Extraneous or spurious peaks also can originate from contaminants in the ionization chamber. In addition, about 30% of EI spectra lack a molecular ion in 70 eV spectra due to high molecular weight or thermal instability. Here structures can be assigned mistakenly to a homologue or a degradation product. Most GC-MS instruments offer chemical ionization (CI) sources and are easily converted to CI operation but CI is underutilized as a routine technique for qualitative analysis at this time.

The present paper describes stepwise development of the new database software, and provides a brief overview of its operation. Because of the importance of accurate isotope ratios in application of the MTS Search Engine, the performance of two quadrupole GC-MS instruments, one using isobutane as reagent gas and the other methane, was studied. Finally, the overall qualitative performance and effectiveness of the MTS Search Engine was studied by inputting experimental mass spectra of over 50 test compounds varying in mass from 75 to 375 Da. These compounds included a variety of pesticides, nitrosoamines and other toxic substances and both halogenated and nonhalogenated structures. The performance testing with unit mass resolution spectra represented a significant qualitative challenge.

Experimental section

Modification of the MTS Search Engine

The MTS Search Engine was modified and tested in a stepwise fashion and a series of software versions were developed to check computations and to optimize functionality. The basic operation of the MTS Search Engine requires setting the mass tolerance, entering the experimental mass, and identifying the ion type by clicking check boxes (e.g., MH⁺, MNa⁺, MNH₄⁺), all at the main page. The search results are tabulated with candidates ranked in order of correspondence to the theoretical mass. The search results are displayed in a table with columns listing: compound name, empirical formula, theoretical mass (Da), mass error (ppm) and reference links to the internet.

An isotope calculator was integrated as follows. First, new software was introduced to calculate the theoretical relative intensities (relative to MH⁺) of the A + 1 and A + 2 isotopes based on the tabulated empirical formulas in the database. Individually these are isotope ratios, but multiple isotopes constitute an isotope pattern. The modified database allows input of the experimental isotope intensities in additional fields. As before the default display arranges candidates in order of mass error (ppm). In the new database the list can be resorted according to either the A + 1 or A + 2 isotope abundance absolute error (%). In a subsequent version both the A + 1 and A + 2 isotope ratios were interpreted simultaneously using a new parameter, isotope fit, the mean of the absolute isotope ratio errors subtracted from 1000. Thus, an isotope fit of 1000 represents a “perfect” isotope pattern. This parameter was easier to use and more effective than considering the ratios independently. A later software version took into account the contributions of the isotopes from the adduct ions. This can be negligible in the case of MH⁺ or of minor significance in the case of MNH₄⁺. The value of deconvoluting interfering isotope clusters was recognized after studying CI spectra of halogenated compounds. Therefore, the MTS Search Engine was further modified to allow automatic deconvolution of M⁺ and [M − H]⁺ ions. The intensities of these ions are entered in new fields that appear after clicking boxes labeled “deconvolute M⁺” and/or “deconvolute [M − H]⁺.”

Mass spectrometry

Chemical ionization (CI) GC-MS spectra were determined with two instruments. Isobutane CI spectra were obtained with a Thermo-Finnigan TSQ 7000 mass spectrometer interfaced to a Thermo-Finnigan Trace 2000 GC (Finnigan MAT, San Jose, CA). The GC column was a 30 m J & W Scientific DB-5 capillary (0.25 mm i.d., 0.25 µm film thickness) (Folsom, CA) and an oven temperature program was used: 40 °C (one min), 40 °C to 150 °C at 30 °C min⁻¹; 150 °C to 300 °C at 10 °C min⁻¹ (inlet, 200 °C; transfer line, 300 °C; carrier gas, 1 mL He per min). The mass spectrometer conditions were: source, 140 °C, isobutane pressure, 1300 mTorr, electron energy, 200 eV: electron current, 300 µA; and electron multiplier, 1,700 V. The tandem instrument was operated in the single quadrupole mass scanning mode with a scan range of 70–500 Da. Pesticides (Table 1) as 10 ppm ethyl acetate solutions were analyzed by isobutane CI-GC-MS and these compounds were also analyzed by electron ionization (EI) GC-MS to determine compatibility with GC-MS conditions, and for comparison to conventional library identification procedures.

Table 1 Pesticides analyzed by isobutane GC-MS

Compound	CAS no.	Formula	Monoistopic mass (Da)
Acephate	30560-19-1	C4H10NO3PS	183.01199
Bromacil	314-40-9	C9H13BrN2O3	260.01604
Ethylan	72-56-0	C18H20Cl2	306.09421
Iodophenfos	18181-70-9	C8H8Cl2IO3PS	411.83536
Lethane	112-56-1	C9H17NO2S	203.09800
Linuron	33-55-2	C9H10Cl2N2O2	248.01193
MCPA, isooctyl	26544-20-7	C17H25ClO3	312.14922
Methidathion	950-37-8	C6H11N2O4PS3	301.96186
Metribuzin	21087-64-9	C8H14N4OS	214.08883
Monalide	7287-36-7	C13H18ClNO	239.10769
Phosphamidon	13171-21-6	C10H19ClNO5P	299.06894
Pronamide	2395-05-85	C12H11Cl2NO	255.02177
Propargite	2312-35-8	C19H26O4S	350.15518
Tebuthiuron	34014-18-1	C9H16N4OS	228.10448
Tetradifon	116-29-0	C12H6Cl4O2S	353.88426
Triallate	2303-17-5	C10H16Cl3NOS	303.00182
Triazophos	24017-47-8	C12H16N3O3PS	313.06500
Trietazine	1912-26-1	C9H16ClN5	229.10942

---

Methane CI GC-MS spectra were obtained with an Agilent 5973 MS interfaced to an Agilent 6890 gas chromatograph (Agilent Instruments, Wilmington, DE, USA). The MS operating conditions were: methane pressure (20 psig); electron energy (150 eV); filament current (200 µA); source temp. (250 °C); electron multiplier (1153 V); and scan range (50 to 510 Da). The GC conditions were: column (J & W DB-5.626 column, 30 m × 0.25 mm, 1 micron); inlet (275 °C); helium carrier gas (1 mL min⁻¹); oven temp. program (60 °C, 1 min; 25 °C min⁻¹ to 285 °C, 285 °C for 10 min); and transfer line (260 °C). Thirty four pesticides and environmental pollutants (Table 2) were determined by methane CI GC-MS.

Table 2 Pesticides/pollutants analyzed by methane CI GC-MS

Compound	CAS no.	Formula	Monoisotopic mass (Da)
Atraton	1610-17-9	C9H17N5O	211.14331
Atrazine	1912-24-9	C8H14ClN5	215.09377
Bromacil	314-40-9	C9H13BrN2O2	260.01604
Butylate	2008-41-5	C11H23NOS	217.15004
Chlorpropham	101-20-3	C10H12ClNO2	213.05566
Chlorpyrifos	2921-88-2	C9H11Cl3NO3PS	348.92628
Cycloate	1134-23-2	C11H21NOS	215.13439
Dichlorvos	62-73-7	C4H7Cl2O4P	219.94590
Diphenamide	957-51-7	C16H17NO	239.13101
EPTC	759-94-4	C9H19NOS	189.11873
Ethoprophos	13194-48-4	C8H19O2PS2	242.05641
Methyl paraoxon	950-35-6	C8H10NO6P	247.02457
Metolachlor	51218-45-2	C15H22ClNO2	283.13391
Mevinphos	7786-34-7	C7H13O6P	224.04498
Molinate	2212-67-1	C9H17NOS	187.10308
Napropamide	15299-99-7	C17H21NO2	271.15723
N-Nitrosodiethylamine	55-18-5	C4H10N2O	102.07931
N-Nitrosodimethylamine	62-75-9	C2H6N20	74.04801
N-Nitrosomethylethylamine	10595-95-6	C3H8N2O	88.06366
N-Nitroso-n-butylamine	924-16-3	C8H18N2O	158.14191
N-Nitroso-n-propylamine	621-64-7	C6H14N2O	130.11061
N-Nitrosopiperidine	100-75-4	C5H10N2O	114.07931
N-Nitrosopyrrolidine	930-55-2	C4H8N2O	100.06366
Norflurazon	27314-13-2	C12H9ClF3N3O	303.03862
Pebulate	1114-71-2	C10H21NOS	203.13439
Prometryn	7287-19-6	C10H19N5S	241.13612
Pronamide	23950-58-5	C12H11Cl2NO	255.02177
Propachlor	1918-16-7	C11H14ClNO	211.07639
Propazine	139-40-2	C9H16ClN5	229.10942
Terbacil	5902-51-2	C9H13ClN2O2	216.06656
Terbutryn	886-50-0	C10H19N5S	241.13612
Tetrachlorvinphos	22350-76-1	C10H9Cl4O4P	363.89926
Triadimefon	43121-43-3	C14H16ClN3O2	293.09310
Trifluralin	1582-09-8	C13H16F3N3O4	335.10929
Vernolate	1929-77-7	C10H21NOS	203.13439

---

Background subtracted spectra were obtained and spectrum lists were produced to determine the intensity of MH⁺ ions and isotope peaks and any M⁺ and [M − H]⁺ ions present – the spectra were averages from multiple scans across the chromatographic peaks. Background subtraction helps obtain “pure” component spectra and, presumably, improved isotope data.¹⁸

Availability of the MTS Search Engine database

Interested researchers are encouraged to request a copy of the latest version of the MTS Search Engine database which is available without charge and can be obtained by contacting the corresponding author. The database will be improved by input from users, and this feedback is greatly appreciated.

Results and discussion

Basic operation of the MTS Search Engine with isotope analysis

The MTS Search Engine contains tables with the theoretical masses of known compounds. The experimental mass from the unknown's spectrum is searched against the database generating a list of candidate structures.¹ The mass tolerance is set in accord with the instrument mass accuracy. For accurate mass instruments, such as time-of-flight (TOF) mass spectrometers, the mass tolerance is narrow, e.g., ±0.001 Da, but for unit mass resolution data, the mass tolerance is wide, e.g., ±0.5 Da. Setting the tolerance correctly ensures that all appropriate candidates are retrieved from the database.

Database searches are notated by the spectral ion, the experimental mass and the mass tolerance, e.g., an unknown with MH⁺ of m/z 248.1 determined with a unit mass resolution, quadrupole instrument is searched as MH⁺/248.1 ± 0.5 Da. Identification with accurate mass data is usually rapid because there are only a few (or one) candidates, but the process can be impractically slow with unit mass resolution data. The hypothesis of the current research is that isotope information can be used to filter or rank MTS Search Engine lists speeding identification.

Because the MTS Search Engine paradigm relies on molecular mass information, chemical ionization (CI) mass spectrometry is preferred to electron ionization (EI). CI mass spectrometry employs ion molecule reactions for ionization minimizing fragmentation. In the CI source the reagent gas is at high pressure and is ionized by electron ionization forming reagent ions (e.g., CH₅⁺, C₂H₅⁺, and C₃H₅⁺ from methane or t-C₄H₉⁺ from isobutane) – sample ionization occurs via ion–molecule reactions including proton or charge transfer.^19–21 The principal ions in methane CI are MH⁺, [M + C₂H₅]⁺ and [M + C₃H₅]⁺ while in isobutane CI MH⁺ ions predominate. Other ion–molecule reactions take place including hydride abstraction yielding [M − H]⁺ ions and charge exchange leading to M⁺.

Additional information is associated with the isotope peaks. Each element is composed of isotopes with differing masses – for example, carbon, has an atomic weight of 12.011 g mol⁻¹ due to contributions from ¹²C and ¹³C (1.1% natural abundance). The relative abundances of the A + 1, A + 2 and other isotope peaks are determined by the elemental composition, the A + 1 isotope peak having contributions from ¹³C equivalent to 0.011 × # of carbon atoms, and so forth. Given the empirical formula, the theoretical pattern of the isotopic peaks can be calculated.¹⁹

MTS database search

As an illustration of the database search the isobutane GC-MS spectrum of acephate insecticide (Table 3) is considered. The background-subtracted spectrum has a base peak of m/z 183.92 (MH⁺) and a minor fragment at m/z 142.85. The spectrum list provides the relative intensities of the MH⁺ isotope peaks, 5.64% and 5.09% for A + 1 and A + 2 peaks, respectively. This MTS database search is annotated MH⁺/183.92 ± 0.5 Da (5.64%, 5.09%) where the mass tolerance is 0.5 Da.

Table 3 Experimental isobutane CI GC-MS Spectra

	MH^+	Relative intensity (%)
		[M − H]^+	M^+	A + 1	A + 2
a Base peak m/z 231.13 (C16H23O) – the abundance relative to MH^+ is tabulated.b Base peak m/z 172.03 (C7H13N3S, [MH − CH3–NCO]^+) – the abundance relative to MH^+ is tabulated.c Base peak m/z 271.09 ([MH − HCl]^+) – the abundance relative to MH^+ is tabulated.d ND = Not Detected.
Compound/tR(min)
Acephate (9.04)	183.92	0.238	1.79	5.64	5.09
Lethane (9.00)	204.07	0.99	ND	11.0	5.19
Methidathion (12.0)	302.95	ND^d	3.61	10.6	14.7
Metribuzin (11.0)	215.10	ND	4.14	11.2	5.78
Propargite^a(13.3)	351.12	ND	237	28.1	8.23
Tebuthiuron^b(9.30)	172.03	ND	ND	9.84	4.84
Triazophos (13.0)	314.07	ND	ND	14.9	5.41
Halogenated compounds/tR(min)
Bromacil (11.6)	260.97	ND	4.87	16.5	99.0
Ethylan^c(12.6)	307.03	128	98.0	64.2	26.7
Iodophenfos (12.2)	412.82	ND	ND	10.0	68.6
Linuron (11.4)	249.04	ND	6.07	13.7	63.4
MCPA, isooctyl (12.0)	313.14	ND	23.4	25.7	33.8
Monalide (10.8)	240.12	ND	8.1	17.1	32.74
Phosphamidon (11.0)	300.03	ND	ND	10.3	30.76
Pronamide (10.4)	255.99	ND	3.92	15.7	63.9
Tetradifon (13.9)	354.88	ND	3.20	19.5	134
Triallate (10.5)	303.99	ND	2.33	14.6	96.9
Trietazine (10.3)	230.10	0.44	8.65	13.8	31.9

---

The spectral data are entered manually into the MTS Search Engine at the main page by inputting numeric data in four boxes (Fig. 1): the Search Mass (Da), the A + 1 relative abundance (%), the A + 2 relative abundance (%), and the mass tolerance (Da). Additional information is entered in check boxes that identify the ion type, e.g., MH⁺, MNa⁺, MNH₄⁺, MCs⁺, R₄N⁺ and [M − H]⁻. Four additional check boxes direct the search to tables of halogenated (Cl and Br) compounds, pesticides (only), drugs (only) and a supplemental table of compounds. In the acephate isobutane GC-MS example the MH⁺ box is selected. Were the spectrum of an unknown, of course, it would not be known that the substance was a pesticide or drug. The pesticide or drug check boxes are used if these are the only substances of interest. The halogen isotope clusters of Cl- and Br-containing compounds are readily apparent to the analyst, and again restrict the search to the appropriate lists.

Fig. 1 MTS Search Engine main page for entering mass spectrometry data.

The search is initiated by clicking the “Search” button. In seconds the search is complete and a Microsoft Access dialogue box opens indicating the number of database tests performed (e.g., total tests = 1204) and the number of matches found. In this case there are 14 compounds in the database that have theoretical MH⁺ masses of m/z 183.92 ± 0.5 Da corresponding to monoisotopic masses between 182.4127 Da and 183.4127 Da. Another click and the 14 candidates are displayed in an MTS Search Report window.

The MTS Search Report

The MTS Search Report summarizes the experimental search mass, the corresponding monoisotopic molecular mass, the mass tolerance, the A + 1 and A + 2 isotope abundances and the # of candidates found. A maximum of 25 candidates is displayed in a table with 7 columns providing: compound name, empirical formula, theoretical mass, mass error (ppm), A + 1 and A + 2 abundance errors (%), and the isotope fit (Fig. 2).

Fig. 2 MTS Search Engine Search Report page displaying results of a database search.

The default display lists the candidates in order of agreement with the experimental mass. Clicking the cursor on each column heading (e.g., mass error, A + 1 abundance error, A + 2 abundance error or isotope fit) resorts the candidates. In the default display (Fig. 2), acephate, with a mass error of −544.9 ppm, is ranked #2. Resorting the 14 candidates by either of the isotope abundance errors or isotope fit moves acephate into the #1 position because the experimental isotope data are in good agreement with theory, e.g., the A + 1 abundance error is −2%, the A + 2 abundance error is −1% and the isotope fit is 999.

For computers connected to the internet the final column in the MTS Search Report provides a link to a web address where chemical information is provided. In the case of acephate the link is to Wikipedia (http://en.wikipedia.org/wiki/Acephate) where the structure is provided as well as chemical nomenclature, common names, physical–chemical properties, toxicology and hazard information, and references.

Deconvoluting interfering isotope clusters

While ion–molecule reactions leading to protonation predominate in CI, other ion–molecule reactions occur which can interfere and distort the MH⁺ isotope pattern. In particular, charge exchange leading to M⁺ and hydride abstraction leading to [M − H]⁺ may occur. If these ions are abundant they must be deconvoluted to allow accurate determination of the MH⁺ isotope abundances. In this study chemical ionization with either isobutane (Table 3) or methane (Table 4) lead to formation of the interfering M⁺ and minor [M − H]⁺ ions.

Table 4 Experimental methane CI-GC-MS mass spectra

	MH^+ (m/z)	Relative intensity (%)
		[M − H]^+	M^+	A + 1	A + 2
a ND = Not Detected.
Compound/tR(min)
Atraton (10.30)	212.1	19.3	12.1	12.3	0.893
Butylate (8.72)	218.1	14.4	3.19	13.5	5.52
Cycloate (9.96)	216.1	38.0	10.7	13.4	5.33
Diphenamide (12.23)	240.1	4.77	3.18	18.2	1.74
EPTC (8.31)	190.1	9.81	2.26	11.3	5.07
Ethoprophos (9.90)	243.0	6.11	2.80	11.2	9.82
Met. Paraoxon (10.85)	248.0	0.426	6.75	9.02	1.55
Mevinphos (8.63)	225.0	0.147	12.0	7.01	1.18
Molinate (9.53)	188.1	21.4	5.44	11.1	5.08
Napropamide (13.40)	272.1	4.92	21.2	18.3	2.15
N-Nitrosodiethylamine (14.58)	103.2	0.156	6.83	5.05	0.295
N-Nitrosodimethylamine (10.00)	75.2	0.138	6.95	2.93	0.233
N-Nitrosomethylethylamine (12.53)	89.1	0.184	6.72	4.05	0.268
N-Nitroso-n-butylamine (22.68)	159.3	1.45	1.18	9.49	0.609
N-Nitroso-n-propylamine (18.78)	131.2	4.44	3.67	7.42	0.459
N-Nitrosopiperidine (19.8)	115.2	0.495	6.74	6.09	0.347
N-Nitrosopyrrrolidine (18.72)	101.2	0.527	7.66	5.22	0.349
Pebulate (8.91)	204.1	13.8	2.50	12.5	5.35
Prometryn (11.58)	242.1	15.2	10.8	13.5	5.03
Terbutryn (11.37)	242.1	34.0	5.81	14.6	5.81
Trifluralin (9.86)	336.1	13.9	8.71	15.5	1.98
Vernolate (8.83)	204.1	13.7	2.44	12.4	5.27
Halogenated compounds/tR(min)
Atrazine (10.46)	216.1	24.4	16.2	14.8	30.1
Bromacil (11.69)	261.0	13.4	3.07	11.6	89.3
Chlorpropham (11.80)	214.0	16.5	8.54	13.4	33.2
Chlorpyrifos (11.85)	349.9	ND^a	1.34	12.2	105
Dichlorvos (7.56)	220.9	ND	4.26	7.29	63.5
Metolachlor (11.89)	284.1	11.9	2.86	17.4	32.3
Norflurazon (15.24)	304.0	36.5	23.2	19.5	29.3
Pronamide (10.65)	256.0	5.61	16.8	24.2	69.9
Propachlor (9.80)	212.0	5.05	4.16	13.3	32.3
Propazine (10.46)	230.1	11.0	16.4	16.3	31.3
Tetrachlorvinphos (12.94)	364.9	ND	ND	11.9	131
Triadimefon (12.01)	294.1	0.82	0.27	17.9	32.9

---

The MTS Search Engine incorporates calculations that automatically deconvolute the isotope clusters, and the calculator only requires information on the relative abundances of the M⁺ and [M − H]⁺ ions. The methane CI mass spectrum of the pesticide norflurazon is illustrative. The base peak is the MH⁺ ion and the relative abundances of M⁺ and [M − H]⁺ ions are 23 and 37%, respectively (Table 4). Because norflurazon (MH⁺, m/z 304) is a chlorine-containing compound, the m/z 306 ion contains significant contributions from the MH⁺ A + 2 isotope as well as the [M − H]⁺ A + 4 isotope. Similarly, the MH⁺ A + 1 isotope and the M⁺ A + 2 isotope overlap. The experimental methane CI spectrum of norflurazon has a cluster of ions at m/z 302 to m/z 308 with intensities as follows: m/z 302 ([M − H]⁺, 37%), m/z 303 (predominantly M⁺, 23%), m/z 304 (predominantly MH⁺, base peak) and associated isotopes. The experimental isotope abundances for A + 1 and A + 2 are entered as before. Selecting the “deconvolute M^+” and “deconvolute [M − H]⁺” boxes opens fields for inputting the M⁺ and [M − H]⁺ relative abundances. Without deconvolution the isotope fit for norflurazon is poor, only 975, but deconvoluting increases the isotope fit to 999. The corrected A + 1 and A + 2 abundance errors are only −2% and −1%, respectively. Without deconvolution it is common to see large errors in the observed MH⁺ A + 1 intensities for halogenated compounds as discussed later.

Testing database performance: identifying volatile and semivolatile compounds with single quadrupole MS data

In this study the spectra of 52 compounds were determined using two quadrupole GC-MS instruments, one employing isobutane as a reagent gas and the other methane. The model compounds included pesticides and toxic substances covering the mass range of ∼75 Da to 375 Da and included various chemical classes. The CI spectra were examined to obtain the mass of the MH⁺ ion as well as the relative abundances of A + 1 and A + 2 isotopes and interfering M⁺ and [M − H]⁺ ions. These experimental data were used to evaluate the qualitative performance of the database.

Chemical ionization spectra

Molecular mass information was readily obtained by examination of most of the CI mass spectra. Almost all of the compounds produced MH⁺ ions as the base peak in isobutane spectra, but this was not always the case. The exceptions were bromacil (MH⁺, 40% relative abundance), iodophenfos (55%), pronamide (93%), and tetradifon (93%). Identification of the MH⁺ ion in methane CI was aided by the accompanying [M + C₂H₅]⁺ ion – the doublet with 28 Da spacing was clear in all cases confirming the identity of the MH⁺. Additional research is needed to investigate gentle or soft ionization techniques to enhance molecular weight information for some compound classes. The present study made no particular effort to optimize the instrument operating conditions for this purpose.

Mass accuracy

The MH⁺ mass errors for the 52 test compounds varied from −3 ppm (triadimefon, Table 6) to as much as 1947 ppm (N-nitrosodimethylamine, Table 6) with a standard deviation (SD) of 456 ppm. This mass accuracy is within the typical performance specifications for quadrupole instruments. Much smaller mass errors are possible with accurate mass instruments, which allow smaller tolerances for database searching.

In this study the number of candidates retrieved in MTS database searches averaged 13 compounds (SD = 7.6). For halogenated compounds the number of candidates was smaller averaging 7.6 (SD = 2.9) while compounds with no halogen retrieved 18 candidates (SD = 7.2). This difference is attributed to the sizes of database tables. The process of evaluating long lists of compounds is time consuming and impractical, and the lack of a rapid and reliable method for evaluating structures can be an impediment. In this case the correct structure was ranked #5 on average (mean = 4.7, SD = 5.4). Thus, the use of MH⁺ mass as a sole criterion without orthogonal information is inadequate for identifying compounds using the 0.5 Da mass tolerance. The method generally requires either high mass accuracy data or additional orthogonal information.

Isotope measurement accuracy

A + 1 and A + 2 isotopes were measured accurately in this study. For the 29 test compounds containing no halogen the theoretical (or calculated) A + 1 isotope ratios varied between 3 and 22% (11.2% on average). The experimentally determined A + 1 isotope ratios were in excellent agreement (Fig. 3) with an average measurement error of 4.2%. For these compounds the theoretical A + 2 isotope ratios ranged from 0.2 to 14%. The measurement error for A + 2 isotope ratios was slightly higher, 7.2%. The correlation coefficients (R²) between theoretical and experimental isotope ratio were 0.9461 (A + 1) and 0.9941 (A + 2) (Fig. 3 and 4).

Fig. 3 Relationship of theoretical and experimental A + 1/A isotope ratios for nonhalogenated compounds determined by CI GC-MS.

Fig. 4 Relationship of theoretical and experimental A + 2/A isotope ratios for nonhalogenated compounds determined by CI GC-MS.

For halogen-containing test compounds the A + 2 isotope ratios were determined with similar accuracy. For 29 test compounds the theoretical A + 2 isotope ratio varied between 33 and 135%. The average A + 2 measurement error was 4.1% and the correlation coefficient (R²) was 0.9949.

A + 1 isotope ratios for the halogen-containing compounds, however, could not be determined accurately without deconvolution of the interfering M⁺ and [M − H]⁺ isotope clusters. In particular, the M⁺ A + 2 isotope peak causes large positive errors in the MH⁺ A + 1 isotope. The observed error is a function of both the M⁺ abundance and the halogen composition of the compound. For example, in the monobromo compound, bromacil, a small M⁺ ion (3.07% relative intensity) results in a +54% error in the measured A + 1 isotope ratio. Without deconvolution the measurement error averaged 25% – after deconvolution the error was only 2.7%. The importance of deconvolution is seen graphically in Fig. 5 and 6 where the theoretical and experimental A + 1 isotope ratios are plotted and the correlation coefficients are 0.6445 and 0.974 before and after deconvolution, respectively. Deconvolution was used for all compounds whenever M⁺ or [M − H]⁺ ions were present, although the correction was only required for accurate determination of the A + 1 isotope ratio in halogenated compounds.

Fig. 5 Relationship of theoretical and experimental A + 1/A isotope ratios for halogenated compounds determined by CI GC-MS.

Fig. 6 Theoretical and experimental data for A + 1/A isotope ratios for halogenated compounds determined by CI GC-MS with deconvolution of M⁺ and [M − H]⁺ isotope clusters.

Integrating isotope information in the MTS database search

Tables 5 and 6 demonstrate the effectiveness of using isotope information with the MTS Search Engine. A good illustration is found in the case of atraton where the experimental MH⁺ ion, m/z 212.1 (Table 4) search is annotated MH⁺/212.1 ± 0.5 Da (12.3, 0.893). This search retrieves 34 candidates and, based on mass accuracy (−242 ppm), atraton is ranked #31. The default candidate display ranks according to mass accuracy, but clicking on headings labeled “A + 1 Abundance Error (%)” or “A + 2 Abundance Error (%)” or “Isotope Fit” resorts the list accordingly – atraton is then ranked #12, #1 and #1 (Table 6), respectively.

Table 5 Ranking MTS Search Engine candidates for isobutane CI GC-MS data based on isotope ratios and patterns^a

	MH^+ mass		A + 1		A + 2		Isotope fit
	Error (ppm)	Rank	Error (%)	Rank	Error (%)	Rank	Value	Rank
a Spectra were corrected for interference due to observed M^+ and [M − H]^+ if the compound contains halogen or the relative intensity of the interfering ion is >5%.b Includes both general and pesticide table.
Compound (candidates)
Acephate (14)	−545	2	−2	1	−1	1	999	1
Lethane (12)	−176	4	−2	1	−3	4	998	1
Methidathion (40)^b	−65	1	7	1	2	1	995	1
Metribuzin (24)	16	1	0	1	13	8	993	1
Propargite (24)	−123	9	1	1	41	3	979	1
Triazophos (17)	−9	1	−2	1	−10	1	994	1
Halogenated compounds (candidates)
Bromacil (6)	−207	2	7	2	1	1	996	1
Ethylan (perthane) (5)	−235	2	−1	1	9	1	995	1
Iodophenfos (8)	−56	1	3	1	−2	2	997	1
Linuron (11)	82	1	−6	1	−4	1	995	1
MCPA, isooctyl (4)	−55	1	−2	1	−2	1	998	1
Monalide (8)	19	1	−1	1	−3	1	998	1
Phosphamidon (7)	−156	3	−12	3	−10	1	989	1
Pronamide (11)	−155	4	−2	1	−3	2	997	1
Tetradifon (5)	−34	1	7	1	−1	1	996	1
Triallate (8)	−65	1	0	1	−5	1	997	1
Trietazine (13)	−75	1	−5	3	−3	1	996	1

---

Table 6 Ranking MTS Search Engine candidates for methane CI GC-MS data based on isotope ratios and patterns^a

Compound (candidates)	MH^+ mass		A + 1		A + 2		Isotope fit
	Error (ppm)	Rank	Error (%)	Rank	Error (%)	Rank	Value	Rank
a Spectra were corrected for interference due to observed M^+ and [M − H]^+ if the compound contains halogen or the relative intensity of the interfering ion is >5%.b Includes both general and pesticide tables.c Metolachlor, propachlor and delachlor are isomers and not distinguished.d Secbuthylazine, terbuthylazine, trietazine and propazine are isomers and not distinguished.
Atraton^b (34)	−242	31	4	12	5	1	995	1
Butylate (13)	−266	11	0	1	3	2	998	1
Cycloate (12)	−196	6	0	1	−1	1	999	1
Diphenamide (12)	−162	4	1	1	1	1	999	1
EPTC (12)	−140	3	1	1	−1	1	999	1
Ethoprophos (23)	−265	6	5	3	2	2	996	2
Methyl paraoxon (21)	−131	1	−4	1	−3	2	996	1
Mevinphos (20)	−236	2	−13	1	−20	12	984	1
Molinate (7)	−58	2	−1	1	−1	1	999	1
Napropamide (17)	−240	7	−5	3	1	5	997	2
N-Nitrosodiethylamine (18)	1106	7	−4	1	−6	1	995	1
N-Nitrosodimethylamine (12)	1947	8	−3	1	0	1	998	1
N-Nitrosomethylethylamine (19)	324	8	−2	1	1	1	999	1
N-Nitroso-n-butylamine (18)	950	5	−2	1	−2	1	998	1
N-Nitroso-n-propylamine (23)	627	8	−1	1	4	1	998	1
N-Nitrosopiperidine (17)	989	7	−4	3	−6	1	995	1
N-Nitrosopyrrolidine (17)	1284	7	0	1	12	7	994	1
Pebulate (12)	−208	10	1	1	2	2	998	1
Prometon (26)	−297	22	−6	2	37	18	979	16
Terbutryn/prometryn (11)	−182	8	6	3	12	1	991	1
Terbutryn/prometryn (11)	−182	8	−2	1	−3	1	998	1
Trifluralin (14)	−51	1	0	1	2	1	999	1
Vernolate (12)	−208	10	0	1	1	1	999	1
Atrazine (4)	−7	1	−1	1	−1	1	999	1
Bromacil (6)	−92	2	−6	2	4	1	995	1
Chlorpropham (7)	−298	4	−1	1	5	1	997	1
Chlorpyrifos (8)	−98	2	−3	1	2	1	998	1
Dichlorvos (4)	−244	1	−1	1	−3	1	998	1
Metolachlor^c (9)	−147	3	1	1	−2	1	999	1
Norflurazon (8)	−153	5	−2	1	−1	1	999	1
Pronamide (11)	−116	4	2	2	9	1	995	1
Propachlor^c (3)	−399	2	−5	2	−2	1	997	1
Propazine^d (13)	−75	1	−2	1	−2	1	998	1
Tetrachlorvinphos (11)	−20	1	10	2	−20	1	985	1
Triadimefon (6)	−3	1	8	2	−3	2	994	2

---

Another example is methidathion whose experimental spectrum is found in Table 3. The database search, MH⁺/302.95 ± 0.5 Da (12.3, 0.893), retrieves a candidate list with 40 compounds (Table 5), each within the acceptable mass window. In this case the mass accuracy is excellent (−65 ppm) but none of the 40 candidates can be excluded based on mass accuracy. The isotope ratio information, however, singles out methidathion as the most likely structure.

Isotope data were consistently effective when used in conjunction with the MTS Search Engine. When candidate compounds were resorted according to the A + 1 or A + 2 isotope abundance errors, the correct candidate was frequently (86% of the time) in the first or second rank. In the case of the A + 2 isotope abundance error the correct candidate appeared in the first or second rank in 87% of cases. Combining the information from both A + 1 and A + 2 isotopes in the isotope fit parameter was more effective than considering either isotope individually. When using the isotope fit parameter the correct compound was ranked #1 92% of the time.

The average isotope fit for compounds not containing a halogen was 996 ± 4.6 where an isotope fit of 1000 corresponds to a “perfect” isotope pattern. For both methane and isobutane CI spectra for twenty halogenated compounds deconvolution increased the isotope fit from 978 ± 30 to 997 ± 1.6.

Conclusions

The present study describes improvements to the MTS Search Engine, a spectrumless database for use in qualitative mass spectrometry. The new database incorporates software for integration of isotope information based on the A + 1 and A + 2 isotope ratios. The database also automatically deconvolutes interfering isotope clusters, and deconvolution is shown to be particularly important for accurate determination of the A + 1 isotope ratio in halogenated compounds. The MTS Search Engine, formerly used exclusively for nonvolatile analytes determined by LC-MS, has been shown to be applicable to the identification of volatile and semivolatile unknowns determined by chemical ionization GC-MS as well. The enhanced effectiveness of database searches was demonstrated by rapid identification of low-molecular-weight, volatile unknowns from unit mass resolution spectra. The integration of orthogonal isotope information in the MTS Search Engine similarly enhances the qualitative performance with API-LC-MS spectra and high mass accuracy data.

Acknowledgements

The following staff of the California Department of Public Health contributed to the development of the MTS Search Engine: E. Chandrasena, J. Chithalen, P. Hill, P. Jayalath, M. McKinney, H. Okamoto (deceased), K. Perera, D. Rajapaksa, R. Roehl. D. Wijekoon and D. Xu. This research was supported, in part, by the federal Centers for Disease Control and Prevention of the Department of Health and Human Services and the U. S. Environmental Protection Agency. Mention of trade names is incidental and does not constitute endorsement.

References

1. W. Liao, W. M. Draper and S. K. Perera, Anal. Chem., 2008, 80, 7765.
2. R. J. B. Peters, J. C. W. Rijk, J. E. Oosterink, A. W. J. M. Nijrolder and M. W. F. Nielen, Identification of Unknown Residues Using Bioassay Directed Fractionation, UPLC/TOFMS Analysis and Database Searching, Rikilt – Institute of Food Safety, Wageningen University and Research Centre, Wageningen, The Netherlands, Report 2009.013, November 2009.
3. J. L. Little, C. D. Cleven and S. D. Brown, J. Am. Soc. Mass Spectrom., 2011, 22, 348.
4. M. Zedda and C. Zwiener, Anal. Bioanal. Chem., 2012, 403, 2493.
5. M. Kraus, H. Singer and J. Hollender, Anal. Bioanal. Chem., 2010, 397, 943.
6. S. Ojanpera, A. Pelander, M. Pelzing, I. Krebs, E. Vuori and I. Ojanpera, Rapid Commun. Mass Spectrom., 2006, 20, 1161.
7. A. Kaufman and P. Butcher, Rapid Commun. Mass Spectrom., 2006, 20, 3566.
8. S. Laks, A. Pelander, E. Vuori, E. Ali-Toppa, E. Sippola and I. Ojanpera, Anal. Chem., 2004, 76, 7375.
9. A. Vonaparti, E. Lyris, Y. S. Angelis, I. Panderi, M. Koupparis, A. Tsantili-Kakoulidou, R. J. B. Peters, M. W. F. Nielen and C. Georgakopoulous, Rapid Commun. Mass Spectrom., 2010, 24, 1595.
10. E. M. Thurman, I. Ferrer and A. R. Fernandez-Alba, J. Chromatogr., A, 2005, 1067, 127.
11. I. Ferrer, A. Fernandez-Alba, J. A. Zweigenbaum and E. M. Thurman, Rapid Commun. Mass Spectrom., 2006, 20, 3659.
12. S. Kern, K. Fenner, H. P. Singer, R. P. Schwarzenbach and J. Hollender, Environ. Sci. Technol., 2009, 43, 7039.
13. T. R. Sana, J. C. Roark, X. Li, K. Waddell and S. M. Fischer, J. Biomol. Tech., 2008, 19, 258.
14. F. W. McLafferty, D. A. Stauffer, S. Y. Loh and C. Wesdemiotis, J. Am. Soc. Mass Spectrom., 1999, 10, 1229.
15. F. P. Abramson, Anal. Chem., 1975, 47, 45.
16. W. G. Mallard and J. Reed, Amdis User Guide, U. S. Department of Commerce, National Institute of Standards and Technology (NIST), Standard Reference Data Program, Gaithersburg, MD, USA 20899, http://chemdata.nist.gov/mass-spc/amdis/docs/amdis.pdf, accessed 10 November 2012.
17. T. Alon and A. Amirav, Rapid Commun. Mass Spectrom., 2006, 20, 2579.
18. W. G. Pool, J. W. Leeuw and B. van de Graaf, J. Mass Spectrom., 1997, 32, 438.
19. F. W. McLafferty and F. Turecek, Interpretation of Mass Spectra, University Sciences Books, Sausalito, CA, 4th edn, 1993.
20. P. Crews, J. Rodriguez and M. Jaspars, Organic structure analysis, Topics in Organic Chemistry, Oxford University Press, New York, 1998.
21. J. Barker, Mass Spectrometry, John Wiley and Sons, New York, 2nd edn, 1999, ch. 2.

---