unknown compounds identification based only on MS spectrum

[semia]

Hi everyone,

I've some questions related to compound identification based only on MS spectrum. I'm having TQ-LC-MS and everybody's saying that I can identify unknown compounds from biological (plasma, serum, urine) samples (and other samples like: tissues, sweat,....hair). Is it possible? because I'm not having a spectral library like in GC-MS. Don't I need C13 and H MNR or IR spectrum?
I really can't understand. I thought that TQ is used in general for samples where I know what to expect in sample and it is used for quantitation that's why is a mass selective detector. Am I wrong? Could you please explain it to me?

Thank you very much,

[sepscientologist]

Good luck with that! You may pick up MW via an initial scan and then do another experiment where you select the
molecular ion and fragment it and look at fragments. I have found that identifying unknowns by LCMSMS is a lot more difficult than
most people expect. When I say most people I mean most people who don't actually do it but have an idea of what it is.

[Don_Hilton]

Workign out structures without a library is a lot of work with EI spectra. Solvign a strucutre is all about information. The less the information you have the more difficult the job.

If you have a reaction mixture and several peaks, you can use the information about the mixture and your knowledge of chemistry to make some intellegent guesses. But for a total unknown the information has to come from somewhere else.

[lmh]

There are a few LC-MS MS2 libraries available (my favourite is Metlin from the Scripps institute). Identifying without a library is a nightmare. If you get a tentative ID, you'll probably need to get the authentic standard and demonstrate by comparison of RT and MS2 that it really does match. I think I've got lucky at least twice in the last decade...
Again, a big question in identification of unknowns is how unknown it is. If it's a common small chemical that's well-known but you don't happen to have its spectrum, you might manage. If it's in a library, you might manage. But if it's a genuine complete unknown, and a great big wobbly molecule for which great chunks fall off as big neutral losses, you're up against a horrendously difficult task.

Most people who claim to be able to identify things from MS alone are either good at bluffing, or work on a small subset of closely related molecules (and they really mean that given something related to what they know, then drawing on a lifetime's experience, they can deduce fairly close relatives).

Have you at least got some access to accurate mass? An empirical formula, or the ability to search libraries by exact maqss, will help a lot.

[unmgvar]

if you have absolutely nothing to go with. and no idea what you are looking for then how do you know on what to look for in the chromatogram to begin with? and how do you know what to separate from what?

and with no idea of anything, then yes NMR and FTIR can generally with a combine MS spectrum help you get to the molecule.
but in a chromatogram filled with unknowns, that is a lot of work.

[MaryCarson]

Interesting thread. It should probably be moved over to hyphenated techniques, though.

[CE Instruments]

Complete unknowns , unlikely. If you know enough about what you are looking at you should be able to make accurate guesses regarding the likely compounds and check by running a standard. If not you need accurate mass and should use a QTOF or other high res system,

[Jetjamnong]

Hi semia
I am working on the toxicology lab. I using LC-QTOF instrument for identify unknown (common drugs and drugs of abuse) in human urine, serum or plasma and whole blood with the following step.
1. enzymatic HY for urine and SPE or LLE cleanup.
2. Dry extract sample and reconstitute with mobile phase
3. inject onto LC-QTOF (single stage MS) with +ESI and/or -ESI (most in +ESI) with run time 25 min.
4. search target compound on our database (>1000 entries) with the following criteria, RT, mass accuracy/ mass error and Isotopic pattern
5. When drug(s) present, then re-extract and inject againt with LC-QTOF MS by MRM method to confirm the sample.
-----
6. When unknown peak appear on TIC or BIC, use smart formula to generate chemical formula of that unknown and search compound in drugbank (http://www.drugbank.ca) and/or wikipedia or other sources for preliminaly information.
7. For the urine sample is easier, check it metabolite (M1,M2,M3...as Desalkylation, Hydroxylation, Glucuronidation as posible)
8. Geting Reference material from any sources such as chemical suppliers or from drugs store
9. Analysis referenece material and if it not present the parent compound, it is big problem (but you can use excretion urine) for furture experiment.

Jetjamnong

[Peter Apps]

I would argue that this process does not identify unknowns, it recognises and confirms the presence of target compounds that are suspected to be present (and very sensibly uses selective sample prep and expected retention times to increase the chances of finding them if they are there).

This analytical problem is very different from assigning a chemical name (and structure) to a compound that nobody has ever identified before, or even from identifying e.g. a contaminant whose likely chemical nature is unknown, but which might be a previously known chemical.

There was another thread about this a couple of months ago.

Peter

[Peter Apps]

The other thread is at: viewtopic.php?f=3&t=15467&start=15

Scroll to the end of the first page for the part that is most relevent to the current discussion.

Peter

[tlahren]

I'm having TQ-LC-MS and everybody's saying that I can identify unknown compounds from biological (plasma, serum, urine) samples (and other samples like: tissues, sweat,....hair).

I would say these people ("everybody's") don't know what they are talking about. Right now, we are trying to identify metabolites of small molecules (cyclohexyl phenyl ketone). We have a good idea already of what the basic structures look like but even with combine LC-MSMS, GC-EIMS, GC-PCIMS and GC-HRMS it is impossible to tell exact structure of isomers. The high mass resolution/accuracy is the best option (in my opinion) and combine with EIMS data is the closest you can come to unknown identification with MS only. An even then it is nearly impossible to distinguish isomers. If you only have LC-triple-quad-MS I think it will be impossible to identify an unknown analyte with any certainty. It would be a "best-guess" based on molecular ion and possible likely (expected) fragments in MSMS.

Ty

[yangz00g]

Interesting topic.

In my (and many other) opinion, there are two kinds of unknowns:

1) known unknowns: you are looking for known compounds but you don't know if those compounds are present in sample. Jetjamnong's approach is clearly for this type of unknowns. (but remember, no mass library can include all the compounds known on earth)

2) unknown unknowns: a novel compound. In this case, NMR is necessary

I believe that most persons here know that MS cannot easily distinguish isomers (some may argue H/D MS can)

[trackster]

This is a very interesting discussion. I agree and disagree with some comments posted. I have been fortunate to do primarily "unknown" identification for the last couple decades. For my part, I use primarily LC/MS and GC/MS, but do work with an NMR expert and occassionally we get some useful information from IR. So, there are various types of unknowns - including, but not limited to: - compounds where one knows somethings about the chemistry (e.g. synthetic impurities/side products, related substances or compounds isolated from a related compound or plant material which has historical data), or knows something about the structure (degradants, etc), and types of unknowns where one knows nothing about the structure at all (contaminants, etc fit into this group). The last one is the hardest, but we have had success at such identifications at times. It requires some luck for sure, and to expect it from just triple quad data routinely - not reasonable. You may figure out some, but will require much work and time along with that luck.

For related substances it helps to look at the MS2 fragmentation of the parent compound and other related compounds to study how they fall apart. This can sometimes help greatly. if you can get accurate mass data on the unknown (TOF, QTOF, Orbitrap, etc) or on the unknown and product ions (fragments) that can help solve the puzzle greatly too. If you have access to an Ion Trap MS that can be more helpful as the fragmentation can be more limited to primary fragments and you can get several generations of MS (MS3, 4, 5) and look at a family tree of fragments. If you combine that with accurate mass (e.g. as in an LTQ-Orbitrap) then you can get this family tree of fragment ions with formulas which is very powerful.

all of this depends on how large these molecules are too. And as others have pointed out, you cannot tell much about stereochemistry of the compounds.

Suffice it to say that it is a very difficult task, often not possible with just QqQ data. when possible, would require a lot of work, time, and luck.

May be a good time to let them know you need a more powerful instrument to do this. I would recommend an LTQ-orbitrap as first choice, a QTOF second, but these will cost a bit and still require some experience and work.

Best of luck

Randy

[james little]

We routinely idenfity "known unknowns" with accurate mass GC-MS and LC-MS data.

Summarized the work in two articles for JASMS. See prepress links below:


Free with Chemspider.

http://littledomain.com/james/chemspide ... _ver_6.pdf

Not so free with SciFinder, but more useful than Chemspider in many applications..

http://www.littledomain.com/james/files ... rticle.pdf

More info on website..

http://littlemsandsailing.wordpress.com/