Tuesday, November 13, 2007

An emailer looks at peak identification

The same emailer who looked at chromatograms with us a few weeks ago has looked at the peak identification question, and finds himself unsatisfied with LNDD's eyeballs.

UPDATE: Retracted due to error. The author concludes:

Yes, nevermind. I thought I had found the simple, but overlooked. It is time for me to shut up and listen.

We leave it around as a reminder to be careful.

Mail #1

If you like, pass on these simple spreadsheets and graphs. It is my understanding that the log of the retention time should be compared on an isothermal GC.

[MORE]

[Ex 84 sample 85424 spreadsheet; and Ex 86 sample 85428 spreadsheet]


'424 matching peaks 10 and 11


'424 matching peaks 11 and 12

'428 matching peak 18 and 19

'428 matching peak 19 and 20


I think these graphs should show a straight line fit through the best peak assignments. If so, it appears that the LNDD peak assignments may be more likely wrong than correct.

Of course, we have already shown a lack of consistency in peak heights between the two machines, and peaks moving around relative to each other.

Time didn't permit further investigation across other samples, but please let me know if this is useful.


Mail #2

Look in Ex84 (References to handwritten page numbers)


LNDD 707

1. Page 70 [LNDD 707] identifies the compounds in Mix Cal Acetate 001C to include 5b Androstanol and 5a Androstanol AC


LNDD 720
LNDD 721

2. Page 83 [LNDD720] (and accompanying GC chart) [LNDD 721] identifies retention times on the GC/IRMS instrument at 1246.9 and 1319.7, which should correspond to 5b and 5a, respectively.

LNDD 705

3. Page 68 [LNDD 705] identifies the retention times on F3. The peak at 1265.6 is ignored. Peaks at 1329.1 and 1359.1 appear to be identified as 5b and 5a, respectively.

4. This is very convincing evidence that the peaks are mis-identified. Notice that if the F3 peaks at 1265.6 and 1329.1 are the real 5b, and 5a peaks respectively, the retention time more closely matches the cal-acetate than does the F3 peaks LNDD appears to call 5b and 5a at 1329 and 1359. Also notice the similarity in the difference between the 5b and 5a peaks.

Thus, if these observations are correct, and apply to all samples, this is very convincing evidence that Landis is truly innocent.

This observation is consistent with your earlier overlays related to the Retention Times discussions, and my graphs from yesterday.

25 comments:

Mike Solberg said...

Huh?

Mike Solberg said...

No, something's not right.

The first one on 707 is 5a Androstanol, the internal standard, not the 5a Androstandiol, the metabolite.

He's got the internal standard and the 5aA mixed up.

syi

m said...

What an idiot! Excuse my French!!!

This guy claims to read chromatographs for a living?

I'm a lawyer and can tell he made a big mistake. There was no switching of peaks.

1. These results are for one of the other stages, not the stage 17.

2. The Cal Mix Acetate contains the IS, and 3 metabolites, 11 ketio, etio, and the 5B, one each for the 3 sample fractions F1, F2, F3,

3. The Cal Mix 1319 rt is the 5B for F3, and the 1246.9 is the etio for F2, NOT THE 5B OR 5A.

4. So for the F3 fraction, the 5B is at 1329 rt for the Landis sample, 1330 rt for the blank urine, and 1319 rt for the Cal Mix. Off by a little from the Cal Mix.

The 5A is at 1360 rt for the Landis sample, and 1360 for the blank urine.

5. There was no switching of peaks. The 1246.9 was not FOR THE 5B, but for the etio.

6. The results are summarized at LNND 724.

7. I'm getting tired of dealing with half baked science. And I'm not even a scientist. I want to leave the science to thoe who know what they are doing.

LET ME REPEAT: THERE WAS NO SWITCHING OF the 5A or 5B PEAKS IN THE LANDIS F3. Even duckstrap doesn't claim that.

Mike Solberg said...

m said::

"What an idiot!"

Hey, it wasn't me!

MR. Idiot.

Larry said...

M, Dr. M-A and Dr. Brenna don't seem to want to post here. I think it's up to us idiots to deal with the science as best we can. I vote for giving emailer a chance to explain.

I mean, in the final analysis, how much do I know about sports law? I sometimes picture Jacobs or Suh reading some of the stuff I've written, and laughing their asses off. If the folks here are nice enough to let me fumble my way around the law, we can be generous with those who might be fumbling their way around the science.

And maybe emailer has a point. Though I'm not following it yet.

blackmingo said...

M-

Regarding your point #3 from yesterday:
"3. Elution as a close adjoining peak will not affect the 5A carbon measurements. That is what Brenna testified and even Meier."

Found those slides from Meier -his powerpoint is in the November 7 major document release:
http://trustbut.blogspot.com/2007/11/major-release-of-hearing-documents.html

In particular, I am taking Meier's slide 13 which is a direct quote from an article by Brenna, which reads:
" Overlaping peaks detected by conventional algorithms are systematically distorted in isotope ratio even for closely matched compounds, though high precision is maintained. Further, small trailing peaks can significantly affect the apparent isotope ratio of the major peak."

Later, on slide 36:
"The GC/MS chromatogram of Floyd’s A sample (Fraction 3) is clearly much more complex than the control urine (BLU) let alone the Mix Cal Acetate mix. In the right flank of 5ßA there is clearly a peak not present in the control urine and the peak tail of 5ßA together with the unknown peak tail into 5αA. The peak start of 5αA at 15.4 min is clearly higher than its peak end at 15.6 min. If peak overlap already occurs in GC/MS it will undoubtedly occur in GC/C-IRMS."

My take from Meier's slides:
1) Floyd's sample has overlapping peaks on GC/MS (which could only overlap to a greater degree on GC/C/IRMS.
2) Overlapping or trailing peaks significantly affect the estimate of the isotope ratio of the major peak.

Based on this, I would say at least Meier would disagree with you, and Brenna might disagree on principle.

At the trial, Brenna's testimony:
(from http://trustbut.blogspot.com/2007/05/hearing-monday-brenner-part-i.html)

Q: What about the peaks following the 5a? Would they have any influence on the 5a peak?

A: No.

Q: What about 171? Would the 8 after the 5a affect the result?

A: No. It might for diagnostic purposes if we thought there was a problem.

Q: 173 -- You don't see them as clearly in the IRMS. Does that suggest anything wrong with the IRMS?

A: No. They detect things different ways, the sensitivity of IRMS is related to the carbon content - this is predictable; the sensitivty of GCMS is sensitive to the molecular structure, and harder to predict. We'll see peaks in GCMS we won't in IRMS. So that doesn't concern me.

Q: Do any of those 8 peaks bear on the reliability of the 5a results?

A: No.

Unfortunately, Brenna is not forced to elaborate on why he does not think the GC/MS peak overlapping with 5A would affect the estimate of its carbon content on IRMS.

To me, looks like we got a disagreement from the experts.

Understandably, Brenna, being on team USADA, is not really explaining his thoughts as clearly as Meier can on his powerpoint, so I don't think we have as clear an opinion from him. Perhaps we can arrange a conference call with the gents.

Anyway, your thoughts are appreciated.

Dan

DBrower said...

The emailer writes:

"I made a big mistake on names---sorry."

I'm awaiting further clarification.

TBV

m said...

Mingo,

Brenna replies to Meier's Slide 13 on redirect, since that was his article that was quoted.

Meier testifies in Slides 20 to 24 that a close peak caused the 5A delta to get larger. On redirect, Brenna testified that to the contrary such a peak caused the 5A delta to get smaller.

I quoted this Brenna language in one of my comments way down below. It's also cited in the majority decision and relied upon.

You should also read Meier's testimony regarding slide 16. He explains there that noise and sloping baselines can in most cases be corrected for.

If I remember correctly, it was only Meier's testimony re: slides 20 -24 that raised a specific issue wrt to the F3 chromatograms. Most of his slide show chromatography testimony was general or referred to other examples.

Ali said...

m,

You should read WMA's slide 16. It covers some simple cases which should clear up any misunderstandings you may have.

Ali

m said...

Ali,

As you know I've already read slide 16.

If you want to point me to any of Meier's testimony in the transcript that specifically discusses problems with the F3 chromatograms, I'll take a another look.

blackmingo said...

M-

"Meier testifies in Slides 20 to 24 that a close peak caused the 5A delta to get larger. On redirect, Brenna testified that to the contrary such a peak caused the 5A delta to get smaller."

I don't how either of these gentlemen could comment on how this peak would effect the 5A delta without knowing the what the close peak is (is it natural or synthetic? What is the close peak's C13 delta?).

thanks

dan

Ali said...

m,

Hmmm, read slide 13, then look at the bottom right diagram of slide 16 (yes, slide 16 - worth careful examination). Now look at any of the F3 chromatograms, even the blanks. Those which exhibit a perfectly vertical line seperating the peaks have had that line retrofitted to the curves to attempt to seperate the peaks. In some cases only a small overlap existed. In other cases, a large overlap existed. For the latter, LNDD have decided to adopt an unusual tactic. In some cases, they went down the "it's caused by a sloping background" route, whilst in other, almost identical cases, they decided to include what had previously been regarded as background and split the peaks with an axe (the vertical lines). I think if you humor me and closely examine the F3s, you'll see what I mean.

He was clearly referring to the F3 peaks with this testimony. He talked extensively about overlapping peaks. The F3s exhibited overlapping peaks. Perhaps he should have spelled it out more clearly but it is not unusual for these academic types to overestimate the abilities of their audience.

Ali

m said...

Ali,

"He was clearly referring to the F3 peaks with this testimony."

Show me where in the transcript. I'm not looking for your interpretation of the F3, but his.

I don't believe he ever specifically referred to the F3, except wrt to slides 20-24. I could be wrong, but you'll have to show me where in his testimony.

Ali said...

m,

Let us not turn a molehill into a mountain. If you don't believe he ever specifically referred to the F3, then why do you imagine he focused on overlapping peaks ? Do you believe that this information was provided as a public service ?. While he happened to be on the stand, he thought he'd throw in a bit of general education on interpretation of IRMS chromatograms for the benefit of the public ?. You're the lawyer, go figure.

Furthermore, I'm not on trial. If what is obvious to me is not obvious to you, that's not my problem. I have already tried to explain it to you and you have rejected my explanation for the rather mundane alternative of "if it wasn't said explicitly, it wasn't said". That's a lawyer's game, not a scientist's. Don't expect me to pursue these pointless word games.

This exercise is an attempt to determine the truth. As much as I'd like to define that outcome, I know that I can not. The spreadsheet confirms my reluctant impartiallity. We've not had any negative feedback on that yet, so I can assume that it is generally accepted as a working model. Both pro and anti Landis camps can use it to pursue their own theories.

Respectfully, Ali

blackmingo said...

M-

From:
http://trustbut.blogspot.com/2007/05/hearing-mon-meier-augenstein-iii.html

(referring to slide 39 of Meier's PPT presentation -available on TBV's 'major document dump' page)

q: slide 39, USADA 348, f3 sample B sample. GCMS
a: yes.

q: USADA 349, disappearing peaks, impact?
a: can't say. it's either part of one peak or the other.

q: 5b tails into 5a.
a: it will shift the delta values reported.

q: symmetry is important?
a: yes, if you don't know what the sample is.

---------------------------------
I think this answers the question as to whether Meier specifically addresses F3 in this regard.

other testimony that day that addresses coelution or overlapping peaks :
---------------------------------
Slide 28

q: why this table?
a: illustration of peak overlap effects on reported deltas.

[ shows -27.4 reported at -32.1 ]

[ shows a fatty acid chromatogram ]

q: what does this show?
a: bad chromatogram, good example of strongly overlapping peaks. This slopes up due to column bleed; there is something else.

q: we've heard if small peaks co-elute, it doesn't matter
a: if you know what you are dealing with, yes, but not if you don't.

[ Landis is focused ]

a: if it's hugely different than your target peak, you can't assume the contribution is minor.

[ slide labelled consequences for delta C13 values. from previous fatty acid slide ]

a: so this compares what you get from purely automatic analysis; we see three injections. Software get's a funny number?

q: isn't it close enough?
a: no!

q: opinion on the chromotography here, and the influence on the results.
a: no confidence. peaks have disappeared, co-eluted, overlapped, long tails downward sloping baselines.

q: A sample F1 USADA 158 what do you see:
a: good, stable baseline, no interference, good separation. daylight between peaks.

q: compare fo USADA 165 and 167.
a: the circled area shows them just about separated in the GCMS, the IRMS shows first peak bleeding into second.

q: affect?
a: will affect C13 numbers for both. The first will look more positive, the latter more negative.

q: USADA 171. circles
a: clearly visible shoulders between 5a and 5b, and 8 minor peaks following.

q: GC/MS chromatogram.
a: 4 peaks not 8 following; disappearing shoulders; where did they go?

q: it's so small it doesnt' matter?
a: can't draw any conclusions, only speculate.

q: good enough?
a: terribly sorry, you don't work on asusmptions.

q: they're all cheaters.
a: they all have rights to a fair hearing and access to the data. I don't know what these are, given the discrepency in the retention times.

Thanks M -I'm really glad I read that testimony for myself.

Dan

m said...

Mingo,

cite the usada page numbers or I can't find it.

Your first quote about the F3 was the testimony that Brenna rebutted in the rebuttle testimony (not redirect as I first stated).

Read the majority opinion. They credited Brenna, who wrote the article that Meier is relying on, not Meier.

As to the other testimony I might go back and look at it if you cite the page numbers. But they don't refer to the F3, as far as I can tell. The F3 is the key evidence against Landis.

daniel m (a/k/a Rant) said...

M,

I believe you'll find the testimony Mingo cites on pages 1199 - 1218 of the PDF file of the official transcripts. That may not be all of his citation, but it's certainly a good portion of it.

blackmingo said...

M-

Thanks Rant. Yes -page 1214 the F3 GC/MS is reviewed and on the next page the GC/C/IRMS. Meier suggests the peak between 5B an 5A dissappears on GC/C/IRMS and one can't say whether it has merged with 5B or 5A. When Brenna is shown the chromatographs on rebuttal (page 1682), he simply says he does not see a problem. Suh does not cross him on this.

Re: Brenna's rebuttal of the leading and trailing peaks affecting the carbon 13 estimate.
Brenna's rebuttal is kind of vague, but suggests that the effect of two peaks overlapping is that the trailing peak's C13 actually increases (becomes less negative (p.1680):

"There's two overlapping peaks. Let
2 me get this right. It showed that the direction
3 would be the opposite of what was quoted, and
4 our study was really quite an extensive study in
5 this regard. And so, were there overlapping
6 peaks, particularly in relationship to the
7 5-alpha, I would have expected it to rise, not
8 fall, compared to -- as an artifact. If it were
9 an artifact, which I don't believe it is."

So, we got two gentlemen discussing what would seem a simple issue to resolve experimentally: If two peaks overlap, what is the effect on the second peak (the one on the right with the longer retention time)?

Brenna seems to say the second peak will see its carbon 13 become less negative, while Meier seems to say that it becomes more negative. Brenna has the paper that's published and quoted by Meier. Meier provides experimental data (slides 30-32) that seem to back his point.

Who's right?

"No more yankie my wankie. The Donger need food. "

Dan

Larry said...

M, I write this with great trepidation, knowing what you think of non-scientists offering science explanations. I'm asking that you don't shoot me for doing this. I'm not trying to be Dr. Brenna here. I'm just trying to state what looks like ought to happen when you measure two overlapping peaks, to advance the discussion and to see what we can figure out. No one should use this as a study guide for their final exam on chromatography. Also M, forgive me for writing this in simple language. I'm perfectly happy to have the more science-oriented tear this post to shreds, so long as I learn something in the process, and do not completely lose whatever remains of my self-esteem. And no matter what happens here, I GET IT, I'm not a scientist and anything conclusive is going to have to come from the scientists.

(whew! If that's not enough of a disclaimer, then I give up.)

Take a look at the earlier discussion at Integration for Idiots III. Without meaning to, we started to address the question of what direction the error would tend to go when two IRMS peaks overlap.

Ali and TBV have taught us that when we're looking at a peak on an IRMS chromatograph, we're actually looking at two peaks, closely superimposed: a C13 peak and a C12 peak. The C13 peak appears on the graph slightly earlier than the C12 peak. To get a picture of what I'm talking about, take a look at the first chart in Figure 0 at Integration for Idiots I.

If you have two overlapping peaks on an IRMS chart, you need to keep in mind that there are actually two C13 peaks and two C12 peaks represented on the chart. Again, try to keep in mind the drawing at Figure 0 I cited above, and imagine adding a second set of C13 and C12 peaks to Figure 0, with one set of peaks slightly overlapping the second. What happens when this overlap takes place? Well, for the first (left-most) set of peaks, the overlap will affect the C12 peak more than the C13 peak, because the C12 peak is slightly to the right of the C13 peak. The affect on the second (right-most) peak should be just the opposite - the overlap will affect the C13 peak more than the C12 peak, because the C13 peak is slightly to the left of the C12 peak.

(I'd add a drawing if I could to make this clear. Maybe we can do this at some point if what I'm saying is not entirely clear.)

Now, if I have this straight, the delta-delta test for measuring for the presence of exogenous testosterone looks to see if the ratio of C13 to C12 is suspiciously low. So, any problem with an IRMS chart that tends to underreport the C13 ratio creates a potential for a false positive reading. Now, the crux of my understanding here is that when two IRMS peaks overlap, then for the left-most peak, we'll tend to measure a higher C13/C12 ratio than we should, because the overlap hides more C12 and less C13. That creates a potential for a false negative. Conversely, for the right-most peak, we'll tend to measure a lower C13/C12 ratio, because the overlap hides more C13 and less C12. This creates the potential for a false positive.

This problem is illustrated in the examples TBV and Ali created in part III of the idiot's guide to integration, see cite above. Look at the examples in figures 11, 12 and 13. As TBV and Ali increase the overlap of the two IRMS peaks shown in these figures, the measured C13/C12 value for the right-most peak gets smaller and smaller, and the possibility of a false adverse finding goes up accordingly. Now, TBV and Ali do not show this in part III, but my guess would be that the measured C13/c12 ratio for the left-most peak would INCREASE in these examples as the overlap increases. (TBV and Ali, is this correct?)

So, this idiot with a small "i" thinks that Dr. M-A probably has it right in this particular argument. Of course, we're talking about very simple examples, and even TBV warned me that in the real world we'd have to deal with a number of other possible variables that could push the test results in other directions.

OK, M. Go ahead and do your worst, only be kind to me! I'm just interested in truth, justice and beauty.

Unknown said...

Re: F3 (blackmingo 6:00pm)

Dr. MA cited specific examples in his testimony.

Brenna states he sees no problem.

The Majority Award relies on the testimony of Brenna and does not credit Dr. MA.

Interesting

DBrower said...

To steal a line from the Marx Brothers, it comes down to who you gonna believe, Brenna or your own eyes?

The Majority wanted to believe Brenna.

And here we are.

TBV

blackmingo said...

TBV, Ali, and the rest y'all,

Just want to say sorry I didn't check out the idiots series sooner (among other things) -I really should have. Another well done for you.

Larry, you're funny. and you write well. thanks.

Dr. J. Butcher -it is very interesting, we need an expert tiebreaker (besides ourselves).


Dan

m said...

Larry,

On the road. Don't know when I can get back to you. Probably next week.

But one question to you. So what was Brenna talking about then, when he said the effect of the peak would be to decease the delta, not increase it. You have to reconcile his testimony with your story.

Larry said...

M, I don't see how I can reconcile my proposed explanation with Brenna's testimony. I don't think Brenna explained why he thought that there would be a decreased delta-delta with an overlapping peak. If I don't know his reasoning, how can I offer a reconcilation?

Probably the better question is, why did Drs. Brenna and M-A disagree on such a fundamental point? Unless we can somehow reconcile their testimony, then the best I could hope to do would be to explain the testimony of one doctor but not the other.

Actually, both doctors agreed that, when you have overlapping IRMS peaks, there will be a tendency for the delta-delta reading on the second peak to err in a particular direction. Yes, they disagreed on the direction! But they did agree that there would be a tendency for error and that this tendency was not random.

I'm offering up for discussion a very simple explanation for why there'd be a tendency for error in a particular direction.

In any event, I'm not trying to be adversarial. I'm not a scientist (as you well know), and it isn't possible that my proposed explanation could be the final word here.

DBrower said...

Hi, y'all, on this overlapping peak thing, go look at Everybody is right, but the results are wrong.

TBV