Friday, December 07, 2007

DailBob on different columns

In a belated comment on Different Columns detailed, DailBob does some 'splainin that seems worth fuller exposure:


Before I start, a reminder that we don’t have an IRMS in our lab, so I’m “projecting” some of what I’ll call “GC 101” identification principles onto the IRMS. I think I’m correct in doing this, because this is essentially what the Landis team did via their retention time arguments. I also vetted my thinking with our Analytical Chem Manager, and he’s in the same place I am.

[MORE]



I first want to provide some history and an example to give you some background on what’s driving my perspective. I’m old enough that I remember getting our first GC-MS in the late ‘80s. The addition of the MS was a game changer, because it added another “layer” of identification, the value of which can’t be understated. Prior to that, you could only identify by elution time, and we were sometimes wrong. Now, we’re rarely wrong. This changed the analytical chemistry world to where I think most Chemists would tell you that they require both for identification, i.e., the MS upped the bar. I’m telling you this, because not having the full mass spectra to aid in identifying something on the IRMS places all the emphasis on elution time.

A few weeks ago, we got a sample in where one of our plants thought they accidentally put glycerin in a batch instead of what was supposed to go in. To determine if this happened, we ran a glycerin standard which eluted at 2.854 minutes. We then injected the sample, where a peak eluted at 2.862 minutes (0.008 minutes or about half a second difference). The mass spectra matched the glycerin standard, so these two pieces of information, taken together, identified the material as glycerin. I’ve provided this example to demonstrate how these machines identify compounds. This is the only way they identify compounds, and the identification is pretty “black or white” – either it’s identified, or it’s not (an exception is isomers, but that’s NA here).

I think the main issue as you move a sample from the GC-MS to the IRMS is maintaining the identity of the compound for which you want to determine the CIR. I say this for a couple of reasons. First, even if the GC-MS and IRMS use the same column, the fact that the columns are two separate physical units can cause differences in elution time, or even peak swapping, if one of the columns is contaminated. Second, as I mentioned, because the IRMS cannot identify the compounds with full mass spectra, the only way of identifying it is by elution time. Consequently, I think it is absolutely paramount that it be tied directly to a known standard, because without it, you have zero “layers” of identity.

I have expressed concern about the different temperature ramps between the two machines, which you have discounted (note: Larry has uncovered information that makes it look like the difference in temperature ramps is intentional, but this doesn’t negate my point). If you look at the Landis F3 from the GC-MS, there is a peak that elutes at about 13.2 minutes. This peak seems to have disappeared in the F3 IRMS. This peak either a) contained no carbon, b) moved due to the change in temperature ramps, or c) indicates a contaminated column. I can make this argument for any peak that is in the GC-MS output that doesn’t appear in the IRMS. If I can't account for a peak, and the peak I'm interested in is not directly identified by being tied to a standard, its identity becomes suspect.

You asked me whether I think the 5aA or P could be misidentified? If you can play a mental game for a moment and adopt the perspective that the only way things are identified are by elution time (minimally) and mass spectra (ideally), I think you can see that, technically speaking (i.e., by standard separation science), they aren't identified at all. What you have done in your example is arrived at the identification by logical deduction. While your logic is excellent, logic is not science. Because of the gravity of this case on a person’s life, I have to have the identification be done by standard separation science, particularly because it's capable of doing so in an incontrovertible way, if it's done carefully and correctly.

You asked me once whether I thought the science told us whether Floyd doped, and I replied that I really couldn't tell with any surety. This is why. While I can't prove he didn't dope, this science isn't good enough, in my view, to prove that he did. When I combine this with a) generally crummy chromatography, b) poor lab practices (e.g. white-out), c) the fact that the blank urine was “positive” (for me, indicating some possible machine bias) and d) the fact that I was influenced by Amory’s testimony, I still have enough doubt that the excellence of your logic doesn't push me over the hump.

17 comments:

Mike Solberg said...

Excellent, dailbob. Thanks.

In my humble opinion, TD2003IDCR is set up with the same scientific reasoning you follow here.

syi

dailbob said...

TBV,
Wow. I wasn't anticipating the "front page." I was just trying to have M understand why I seem so hard-headed. Thanks!

m said...

Dailbob,

It's as if you DIDN'T EVEN LOOK at the graphs. Look again please, and tell me whether the 5A could have switched with any other peak, based on retention times and visual matching? The answer is no. Remember the 5B that elutes right before is anchored because it has the same retention time as the 5B in the reference material (mix cal acetate) and in the blank urine sample. So that means the 5A peak is properly identified because it comes right after the 5B. Moreover, the retention times of the 5A match in the sample and the blank urine sample.

Could it contain something else that co-eluted? That is possible. And that's what your mention of the "disappearing GCMS peak" at 13.2 minutes might refer to.
But that is a question of specificity, not identification. Just because the full mass spectra haven't been produced you are claiming doubt in the identification?

You seem to be talking about speculative possibilities and general principals rather than discussing the specific peaks of interest. You don't even discuss the peaks of interest.

You claim the retention times must match. They cannot match because of the combustion phase.

By the way how can you demonstrate that that peak at 13.2 minutes disappeared in the IRMS? I can't tell that. Look again at USADA 348 and 349 and explain where the peak disappeared. Are you referring to the mini-peak occurring aroung 1320 seconds in the GCMS between the 5A and 5B?

Three questions.

1. What is the probability that the 5A peak switched with some other peak in the GC-IRMS and is mis-identified? My answer is 5% or less.

2. The retention time of the 5B in the sample matched the 5B in the reference sample. If the lab had included the 5A in the reference sample instead of the 5B, and retention times similarly matched would you concede a proper identification?

If yes I don't know how you can avoid conceding that the 5A is properly identified here since the 5B is anchored, unless you claim that the 5A peak switched.

I'm posting my graphs below so you can look again.

m said...

In the graphs below A=5B, B=5A, I=Internal Standard, P=pregnane. The a and d are unidentified small peaks.

Was the 5A identified in the IRMS F3 sample. Yes.


GRAPH OF RRT RESULTS GCMS and GC-IRMS


|||||||||||||||||||||||||||||||||||||||||||||||||||||||||scale

1) GCMS Mix Cal
................I........a.A.B.d......P


2) GCMS F3
................I........a.A.B.d......P


3) GC-IRMS F3
......................I..........a.A.B.d.......P


4) GC-IRMS F3 Blank Urine
......................I..........a.A.B.d.......P


5) GC-IRMS Mix Cal
......................I............A............


The B in 1) and 2) match based on RTs, as well as the A, I and P.

The metabolites in the IRMS graphs in 3), 4), 5) eluted later than in 1) because the IRMS involved a combustion phase and because changed ramp conditions caused the mebolites to elute further apart probably to achieve greater peak separation. And possibly because of different column specs.

The RTs in the graphs in the IRMS all matched. So there is no question that the B in the IRMS F3, must be the same B as in the GCMS F3. Remember it comes right after the A, which we have fixed in the 5) graph.



USADA 149 AND 323 summary refers to match of RTs between GCMS F3 and GCMS Mix as satisfying the ID criterial of TD2003IDCR, with no indication must compare IRMS F3 with Mix.

USADA 185 and 351 summary entitled IRMS analysis records RTs between IRMS F3 and IRMS Urine, as well as the Carbon values. Is this identification or quality control?



For the GCMS F3, we know that the IS, A, B, P in the F3 GC are identified based on their MS and comparing their RRTs with the Mix Cal.

We know that the a A B d appear to follow in the same order in both IRMS F3 and the GCMS F3 and GCMS Mix based on RRTs and height of peaks. We expect a longer RT in the IRMS because of the combustion phase, and expect slightly different RRTs because of the lower ramp tempature and difficulty in making GC conditions identical, so we are not matching RTs.

We know from the GCMS F3 that all of the substances identified in the F3 have to be there and that the A and B have to be in the same order unless they switched position with some other substance.

We know that the A in the F3 IRMS is identified because of the Mix Cal IRMS.

For the F3 IRMS, can the B be anything else than the B? Answer no, unless it switched places with the d or a. There is no evidence that it switched based on the heights of the peaks or any other evidence.

Bill Mc said...

Is anyone else having a problem with M's "sanitized" graphs? They are "simplified" to the point that a lot of significant context is lost. I find them misleading. As for his conclusions, I do not find them persuasive, just more meta-science. I also find his frequent use of the phrase "we know" to be unwarranted, because most of the things introduced with that phrase are not "known" in a scientifically verifiable sense.

m said...

bill mc,

Tell me SPECIFICALLY why they are misleading about any relevant point.

dailbob said...

M,
Yes, I looked at the graphs, and yes, I follow your logic why you believe so strongly that the peak in question is 5aA. I even told you in a different thread that you could well be right (based on the same logic), and that I thought it at least contained 5aA. The point I was trying to make is the way you arrived at that identification was through logical deduction, but that is not how the machines are used to identify compounds (and you are correct these are general identification principals, not specifics). I make this statement based on 27 years of lab experience, as well as having access to a couple of guys with about 40 years of HPLC, GC and GC-MS experience between them.

You took the time to construct a graph in order to demonstrate things you had told me previously. The reason for my detailed response was that I felt I owed you my full thinking on why that wasn’t moving me off of my position, that’s all. Take it for what you think it’s worth.

Regards

m said...

Dailbob,

I'm sorry to say, I don't think it's worth much since you dodge all the issues that are specific to this case and instead rely on general principals, which by your own admission may not be applicable to GC-IRMS.

You say:

"I think it is absolutely paramount that it be tied directly to a known standard, because without it, you have zero “layers” of identity."

In your mind the known standard is the mix cal acetate. Neither the lab nor any of the scientific experts that testified took that position.

But even if it were true, are you saying that if the mix cal acetate had contained the 5A instead of the 5B, and the retention times had matched as they surely would have, that you would be technically satisfied that the 5A had been identified?

Then where would that leave the 5B?

Ali said...

m,

Your arguement appears to be that you've seen the GC-MS chromatograms and 5a follows 5B directly. That's the only positive identification of 5a position, correct? I thnk a number of clear explanations have been put forward as to why things may chage between GC-MS peak position and GC-IRMS peak position, but you reject all of these because you know the peak positions, based on the GC-MS chromatograms. I think evryone agrees that the GC-MS peaks are probably correct.

The problem is that you suggest that using a different process from that used on the GC-MS has no impact on peak positions, so you can go ahead and apply the GC-MS pattern to the GC-IRMS. Different columns, different temperature ramps, these are factors which you say have no impact. That's wrong. There's a peak after the 5B OK, we can all see that. On what basis do you claim it is the 5a? ... because that's where the 5a appeared on the GC-MS. That was a different process. You're comparing apples with oranges. That's most definitely not "scientific".

Ali

Bill Mc said...

M,

You wrote: "Tell me SPECIFICALLY why they are misleading about any relevant point."

I find them misleading because, as I wrote in my earlier comment, most of the contextual data is lost in your simplified graphs. The data that is lost is the data that has been discussed by TBV and others to call the "official" interpretations of the graphs into question. I find TBV's presentations where he uses the actual graphs and highlights specific areas for discussion to be much more appropriate for a scientific dialog. If one's goal is to persuade, as would be the case in a college debate, then the technique of extracting and presenting data in a manner which only supports your contentions is ok, but if the goal is to find the truth, and someone's career hangs in the balance, then your technique is not appropriate. It is sophistry versus science - sophistry deals with winning a point whereas science deals with uncovering the truth, even if that truth is an admission that the data is so muddled that no conclusions can be reliably drawn from it.

bostonlondontokyo said...

I wouldn't say that M is engaged in sophistry - I think he's giving a quick reference to charts to all the people in the debate have seen many, many times. I think his arguments have been very persuasive and numerous over the past month or more, so I wouldn't judge. Besides, anyone can take a screen shot, crop it, toss it in photoshop, draw some circles and write a caption. I'm not saying I don't appreciate when it's done, but some of us simply don't have time for that, especially if Web work is what we do for a living. I don't think M is trying to 'win' anything here - this is internet, folks! Let's not get so serious!

There seem to be some key issues of interpretation between the Landis supporters and those who argue for the arbitration decision that simply cannot be bridged. I seriously doubt there is going to be an 'ah HA!' moment when someone backs down and says 'Yes! I was wrong the whole time!'...

I've always felt that M is trying to establish the identification that has been met by the standards set prior to the arbitration. In another arbitration, or some court in the US, EU or elsewhere, there might be different standards as well. However, we're dealing strictly with THIS set of standards, and if the minimum was met. Both TBV and M have established some solid arguments. If either is looking to win, it ain't gonna happen, not here.

m said...

Bill MC,

I asked for a "specific example" of my why my charts were misleading. You responded with some general gobbydeegook and handwaving. I don't think you understand the science at all.

Bill Mc said...

M,

Anyone willing to understand would see that my point is that the dialog should be conducted using the actual graphs rather than abstracted text diagrams which omit important contextual information such as adjacent and overlapping peaks, sloping baselines, etc., which can, and should, affect the interpretation of the data. As for understanding the science, while I am not a specialist in the area of Gas Chromatography (but then neither are you), I do have enough of a science background to follow the discussions.

m said...

bill mc,

We are not talking about overlapping peaks or sloping baselines here, but retention times and identification. Nothing I left out should affect that discussion.

I am not using these graphs to assess the quality of chromatography, which is a separate issue. For you to mention those issues, indicates that you are not following the discussion here, or just wished to throw out a general false argument.

Bill Mc said...

I fail to see how taking a position that the original graphs should be used instead of text diagrams in the discussion of the "science" involved with the FL case is throwing out a false argument. Any comments I have made about the quality of those graphs, or their ambiguity are also not throwing out false arguments - they are observable facts.

In any case you do not seem open to my views and contending with you is not what this site is about, so I suggest that we just agree to disagree on this issue

the Dragon said...

bill mc,

Your error is not slavishly worshiping WADA World and particularly the scientific genius that is LNDD.

My paraphrase of the "Majority" opinion in regards to LNDD. They can't add nor subtract, do not understand the concepts of multiplication or division, have never heard of decimals nor fractions yet they are the formost authority on diferential calculus.

Regards,

tbv@trustbut.com said...

Hi guys,

Please remember to focus on facts and logical arguments.

There has been some creeping into the land of ad-hominem criticisms, and that isn't helpful.

If someone isn't seeing the point you are making, you can either try again or give up, but blaming the person isn't going to win many converts.

It can be emotionally gratifying to deliver a cutting remark, but it doesn't usually progress discussions.

thanks,

TBV