In a series of posts, we are going to talk about Mass Spectrometry.

1. Introduction-The dif­fer­ent con­fig­u­ra­tions and the Elec­tron Impact process
2. What types of mass ana­lyz­ers are there?
3. What type of detec­tors are there?
4. What types of analy­sis can be done?
5. How do you read the output?
6. How do they come to a qual­i­ta­tive mea­sure using software?
7. How do they quan­ti­tate the results?
8. Do you need chro­matog­ra­phy if you are using Mass Spectrometry?
9. Other top­ics of inter­est about GC-MS

I am being very spe­cific with my lan­guage here when the ques­tion posed is “How do THEY come to a qual­i­ta­tive mea­sure using soft­ware?” and is not the ques­tion of “How SHOULD you come to a qual­i­ta­tive mea­sure using software?”

Let’s look at how THEY do it first.…

In this blog, we have posted on this par­tic­u­lar topic before. There­fore, I would encour­age you to review that post now, then come back to this post. Pro­fes­sor McLaf­ferty of Cor­nell once wrote in his book Inter­pre­ta­tion of Mass Spec­tra, “The mass spec­trum shows the mass of the mol­e­cule and the masses of pieces from it. Thus the chemist does not have to learn any­thing new– the approach is sim­i­lar to an arith­metic brain-teaser.”

Mass Spec­trom­e­try is only com­puter assisted pat­tern recognition

Now, pre­sum­ing you have reviewed the above ear­lier post, we can look at some of the most amaz­ing parts of MS work (at least to me) and see how really sub­jec­tive it really is. It is truly open to interpretation.

When we just sim­ply run the NIST spec­tral library searches, we have some issues of concern.

As it comes to pass that all that they do when it comes to the qual­i­ta­tive mea­sure in GC-MS that is EI-based is per­form com­puter assisted pat­tern recog­ni­tion, then we need to be sure that the stan­dard the unknown is com­pared against is from an unim­peach­able source. Is it a trace­able library of the spec­tra or is it local forced inte­gra­tion that is anec­do­tal in nature and there­fore not inde­pen­dently adjudicated?

To try to solve this uni­for­mity prob­lem and to try to homog­e­nize the stan­dards, Pro­fes­sor Fred McLaf­ferty of Cor­nell Uni­ver­sity in 1995 began to get his col­leagues together and col­lect spec­tra. This resulted in the first EI-based mass spec­trom­e­try library that later was con­verted to what we now enjoy as the NIST/EPA/NIH Mass Spec­tral Library. The most cur­rent ver­sion of the EI Library in the NIST ’08 includes:

• main­lib (main EI MS library)=191,436
• replib (repli­cate spectra)=28,307
• nist_salts (EI Salt Library)=717
• nist_msms (MS/MS) Library=14,802 of which 3,898 are pos­i­tive and 1,410 are negative

On the graphic user inter­face as shown above, there are three very impor­tant places that we want to know about in order to judge the valid­ity of this qual­i­ta­tive mea­sure: the text info for the selected hit spec­trum, the hit his­togram, and the hit list.

All of these three key areas are affected by the par­tic­u­lar type of search algo­rithm that is con­ducted. Is it an iden­tity search or a sim­i­lar­ity search?

• An Iden­tity search is designed to find exact matches of the com­pound that pro­duced the sub­mit­ted spec­trum and there­fore pre­sumes that the unknown com­pound is rep­re­sented in the ref­er­ence library.
• A “Sim­i­lar­ity” search is opti­mized to find sim­i­lar com­pounds and is intended for use when a com­pound can­not be iden­ti­fied by the “Iden­tity” search.

What hap­pens in the real world is that the ana­lyst looks may look at the Hit His­togram and then the Hit List. They gen­er­ally ignore the other screens and rarely look at the Text Info for the Selected Hit Spec­trum, although it is per­haps the most impor­tant window.

• The Hit List has dif­fer­ent columns of infor­ma­tion. The Match Fac­tor is an arbi­trary unit num­ber where a per­fect match is 1000. It is the com­par­i­son between the unknown and the library (direct match). The prob­a­bil­ity value for a hit is derived assum­ing that the com­pound is rep­re­sented by a spec­trum in the libraries searched. It only employs the dif­fer­ence between adja­cent hits in the hit list to get the rel­a­tive prob­a­bil­ity that any hit in the hit list is cor­rect. While many state sci­en­tists dis­count the prob­a­bil­ity value, it is arguably, the most “true” (cor­rect) value that we can use to judge the value of the qual­i­ta­tive judg­ment call. Although there is no writ­ten cri­te­ria that is uni­ver­sal, it is thought that a match score of 950 or greater is con­sid­ered an excel­lent match; 900–950 is a good match; 800–900 is a fair match; less than 800 is a very poor match. The Hit List can be thought of like a ranked top 100 list of com­pounds that the com­puter think the unknown is.
• The Hit His­togram is the num­ber of hits vs. their Match Fac­tors. It is dis­played in the pane located just above the Hit List.
• The “Text Info for the Selected Hit Spec­trum” will give us the name of the com­pound, its diag­nos­tic ions, its CAS reg­istry num­ber, its NIST num­ber (if it is a NIST trace­able spec­trum) and most impor­tantly the source of where the stan­dard comes from.

The NIST Spec­tral Library search is very pop­u­lar. It is a short­cut to think­ing through the fun­da­men­tals that mass frag­men­ta­tion is based upon which is noth­ing more than good old fash­ioned acid-base chem­istry. Exclu­sive reliance upon this sim­pli­fied method of analy­sis as a means of iden­ti­fi­ca­tion can lead to an improper answer. There has always been an upstream-downstream prob­lem with the NIST spec­tral library where over the years infe­rior spec­tra have made their way to the offi­cial NIST library. In fact, Pro­fes­sor Mclaf­ferty noted in his book Inter­pre­ta­tion of Mass Spec­tra “[O]ver the last decade approx­i­mately 60,000 errors have been cor­rected in the ref­er­ence file [of the NIST library].” Just con­sider the below as an example:

At least one orga­ni­za­tion has pub­lished that sim­ple reliance on any mass spec­tral library is insuf­fi­cient in and of itself. SOFT/AAFS Foren­sic Tox­i­col­ogy Lab­o­ra­tory Guide­lines 2006 ver­sion in 2006 in sec­tion 8.2.10 we find the fol­low­ing language:

8.2.10 In rou­tine prac­tice, inter­pre­ta­tion of GC/MS-EI full scan mass spec­tra is per­formed by the instrument’s soft­ware as a semi-automated search against a com­mer­cial or user-compiled library. The qual­ity of the match or “fit” may be aided by the fac­tor that is gen­er­ated, either as a ratio or per­cent­age, where 1.0 or 100% are “per­fect” matches. How­ever, such “match fac­tors” must be used as guides only and are not suf­fi­ciently reli­able to be used as the final deter­mi­nant of iden­ti­fi­ca­tion. Final review of a “library match” must be per­formed by a tox­i­col­o­gist with con­sid­er­able expe­ri­ence in inter­pret­ing mass spec­tra; expe­ri­ence and crit­i­cal judge­ment are essen­tial. Inter­pre­ta­tion, at a min­i­mum, should be based on the fol­low­ing prin­ci­ples:
For a match to be con­sid­ered “pos­i­tive”, all of the major and diag­nos­tic ions present in the known (ref­er­ence) spec­trum must be present in the “unknown”. Occa­sion­ally, ions that are in the ref­er­ence spec­tra may be miss­ing from the “unknown” due to the low over­all abun­dance of the mass spec­trum. If addi­tional major ions are present in the “unknown” it is good prac­tice to try to deter­mine if the “extra” ions are from a co-eluting sub­stance or “back­ground” such as col­umn bleed or dif­fu­sion pump oil. Exam­i­na­tion of recon­structed ion chro­matograms of the sus­pected co-eluting sub­stance rel­a­tive to major ions from the ref­er­ence spec­trum will help to deter­mine this.

[Thank you to Dr. Ste­fan Rose, MD for point­ing out the ref­er­ence to the SOFT/AAFS guideline.]

Early in the same stan­dard we find the fol­low­ing language:

8.2.9 Where mass spec­trom­e­try is used in selected ion mon­i­tor­ing mode for the iden­ti­fi­ca­tion of an ana­lyte, whether as part of a quan­ti­ta­tive pro­ce­dure or not, the use of at least one qual­i­fy­ing ion for each ana­lyte and inter­nal stan­dard, in addi­tion to a pri­mary ion for each, is strongly encour­aged where pos­si­ble. Com­monly used accep­tance cri­te­ria for ion ratios is ±20% rel­a­tive to that of the cor­re­spond­ing con­trol or cal­i­bra­tor. How­ever, it is rec­og­nized that some ion ratios are con­cen­tra­tion depen­dent and that com­par­i­son to a cal­i­bra­tor or con­trol of sim­i­lar con­cen­tra­tion may be nec­es­sary, rather than an aver­age for the entire cal­i­bra­tion. Ion ratios for LC/MS assays may be more con­cen­tra­tion and time depen­dent than for GC/MS and there­fore accept­able ion ratio ranges of up to ±25% or 30% may be appropriate.

How­ever, the ulti­mate ques­tion is: “How SHOULD we come to a qual­i­ta­tive mea­sure using software?”

The answer is sim­ple and old: Acid-base chem­istry approach as artic­u­lated by Drs. Fred W. McLaf­ferty and Fran­tisek Ture­cek in their book Inter­pre­ta­tion of Mass Spec­tra. In this book they set out a “Stan­dard Inter­pre­ta­tion Pro­ce­dure” which pro­vides a use­ful and uni­ver­sal stan­dard of how to use the pow­er­ful tool of Mass Spec­trom­e­try cor­rectly to arrive at a qual­i­ta­tive mea­sure. The “Stan­dard Inter­pre­ta­tion Pro­ce­dure” is as follows:

1. Study all avail­able infor­ma­tion (spec­tro­scopic, chem­i­cal, sam­ple his­tory). Give explicit direc­tions for obtain­ing spec­trum. Ver­ify m/v assignments.
2. Using iso­topic abun­dances where pos­si­ble deduce the ele­men­tal com­po­si­tion of each peak in the spec­trum; cal­cu­late rings plus dou­ble bonds.
3. Test mol­e­c­u­lar ion iden­tity; must be high­est mass peak in spec­trum, odd-electron ion, and give log­i­cal neu­tral losses. Check with CI or other soft ion­iza­tion. (Don’t just rely on the EI-based GC-MS result)
4. Mark “impor­tant” ions: odd-electron and those of high­est abun­dance, high­est mass, and/or high­est mass in a group peaks.
5. Study gen­eral appear­ance of spec­trum; mol­e­c­u­lar sta­bil­ity, labile bonds.
6. Pos­tu­late and rank pos­si­ble struc­tural assign­ments for:

(a) impor­tant low-mass ions series;
(b) impor­tant pri­mary neu­tral frag­ments from M+ indi­cated by high-mass ions (loss of largest alkyl favored) plus those from sec­ondary frag­men­ta­tions indi­cated by CAD spec­tra;
© impor­tant char­ac­ter­is­tic ions.

7. Pos­tu­late mol­e­c­u­lar struc­tures; test against ref­er­ence spec­trum, against spec­tra of sim­i­lar com­pounds, or against spec­tra pre­dicted from mech­a­nisms of ion decompositions.

Although there is not for­mally an eighth step Pro­fes­sor McLaf­ferty reminds us that there is another impor­tant aspect to remem­ber in all of this if we want to make sure that we are report­ing out valid and high qual­ity results with great con­fi­dence. He writes:

Remem­ber that abun­dance val­ues can vary by more than an order of mag­ni­tude between instru­ments; so you must mea­sure a ref­er­ence spec­tra of the pos­tu­lated com­pound under the same instru­men­tal con­di­tions used with the unknown in order to have high con­fi­dence in the answer.

Later he writes:

“The most reli­able match is obtained by run­ning the unknown and ref­er­ence mass spec­tra under closely iden­ti­cal exper­i­men­tal con­di­tions on the same instrument.”

So even the “father” of the NIST library requires there to be ver­i­fi­ca­tion by run­ning ref­er­ence mate­ri­als to con­firm spec­trum are in fact repeat­able under the spe­cific con­di­tions of the par­tic­u­lar instru­ment. Dr. Mar­vin C. McMas­ter has plainly expressed the same notion in the fol­low­ing thoughts:

One of the prob­lems with spec­tral library data­bases is that some of their struc­tures are inac­cu­rate or just plain wrong. The orig­i­nal inter­pre­ta­tion of their struc­tures may have been incor­rect or mis­takes may have been made in enter­ing them. The pre­vi­ous Wiley/NIST spec­tral data­base with 225,000 com­pounds was thought to have up to 8% incor­rect structures.

There are many aspects of GC-MS that can lead to erro­neous iden­ti­fi­ca­tions beyond this upstream-downstream prob­lem such as:

1. Were the tun­ing con­di­tions used to pre­pare the ref­er­ence stan­dard in the library the same or dif­fer­ent than the one that the unknown was run upon?
2. The unknown and the ref­er­ence spec­tra could have been run on a dif­fer­ent type of mass spec­trom­e­ter with dif­fer­ent mass lin­ear­ity. For exam­ple, some of the spec­tra in the library are so old that they were run on mag­netic sec­tor instru­ments rather than quadrupoles.
3. The ref­er­ence stan­dard in the library may not have been from a pure com­pound. If the ref­er­ence spec­tra in the library were run on impure com­pounds, then there will likely be addi­tional frag­ment peaks due to the over­lap­ping and the lack of res­o­lu­tion between compounds.

GC-MS work should be reserved to highly sci­en­tif­i­cally trained cre­den­tialed sci­en­tists who have a fun­da­men­tal and strong under­stand­ing of chem­istry, not some­one who can sim­ply fol­low a pro­ce­dure, right dou­ble click and read a screen with­out giv­ing any thought to it all.