Tuesday, November 06, 2007

Integration for Idiots: Part IV: Three peaks

Part IV: Three Peaks

Series by Ali and TBV

In Part III, we looked at "co-elution" (overlap) of two peaks. Now, we'll see some things that may happen when a third is involved.

Figure 14 reverts to the minor co-elution scenario from figure 11. In addition, a small peak with a o/oo value of -50 is shown. What would the effect be if this peak occurred at the same time as our original peak ?


Figure 15 shows what would happen if the small -50 peak occurred at the same time as our original peak. We show the peak in red, but it doesn't show up that way on the display, only as a higher peak in grey. The measured value is now -32.1. It is really -27.


Figure 16 again reverts to the minor co-elution scenario of figure 11. This time we have a small peak with o/oo value of -25, which is less negative than our peaks of interest. Harmless enough, you may think. What would happen if this peak occurred between our minor co-eluting peaks?


Figure 17 places the small -25 peak between the two peaks with minor co-elution. It’s presence has masked where our original peak begins. The measured value for the peak is now -31.6. The true value remains -27.

This series of figures illustrates how the interference between peaks can dramatically alter the measured o/oo value of a peak. In many of the above cases, it would not be possible to determine the true o/oo value.

In Fig 17, is our peak sitting on a sloping background? No, it isn’t, but it may look like it is in context.

As demonstrated in Figure 9, if it had been genuine linear sloping background, we may have recovered the true o/oo value of -27. Instead, we measured a value of -31.6.

Figures 14 and 15 illustrate the effect of a small but significantly negative peak co-eluting with our original peak. Again, this would not be detectable from the data that is available, the M=44 and M=45 traces.


In Part V, having gotten all this background, we'll look at what Herr Doktor Professor Wolfram Meier-Augenstein was saying in just a few slides of his presentation at the hearing.

12 comments:

Mike Solberg said...

Well, now, this is very interesting!

One of the things some people have argued is that if there was some type of co-eluting peak, it would have to have a lower (more negative) CIR in order to make the target CIR lower/more negative. But you are showing that that isn't true. Very interesting.

So, how did LNDD make decisions about where the peaks should begin and end? Team Landis did pay a lot of attention to that, right? Isn't that what the whole "reprocessing" effort was about?

syi

BustinBilly said...

Yes the reprocessing integration was done automatically by state-of-the-art Masslynx software. Those 5alpha androstandiol-5beta pregnandiol differences were even greater.

The integration for the other seven samples was also done automatically by Masslynx.

Kevin said...

Bustin Billy,
And in the reprocessed samples a number of the urine blanks were also positive for exogenous testosterone, and all of them were badly biased in a direction towards a positive finding. Not a great argument for a robust procedure.
Duckstrap

Mike Solberg said...

So, how does the Masslynx make decisions about where the peaks should begin and end?

syi

BustinBilly said...

Kevin - Dr. Thomas Brenna testified that the B sample 5alpha androstandiol-5beta pregnandiol blank reprocessing number resulted from an error in data transfer. He testified that Dr. Simon Davis agreed to pull the result out. Dr. Brenna was surprised to see the number in court.

Mike - I don't know how the Masslynx algorithms work. All parties appeared to have confidence in the software.

tbv@trustbut.com said...

Billy, you said:

Yes the reprocessing integration was done automatically by state-of-the-art Masslynx software. Those 5alpha androstandiol-5beta pregnandiol differences were even greater.

The integration for the other seven samples was also done automatically by Masslynx.



Thanks! As with Larry's comments yesterday, you have anticipated the point to be made in Part V. Well done.

Timm said...

I thought the samples were processed by eye with the ladies "using their experience"? Is the Masslynx software the state of the art program that they were running on that state of the art OS that IBM discontinued a decade or so ago? Maybe I am confused!

hatefrench said...

bustinbilly, or anybody who knows,

could you please give the reference of Brennas testimony, I cannot recall that part and would like to read it. Thanks

tbv@trustbut.com said...

HF, go to the transcripts page, where the index shows Brenna's direct on the 14th at PDF page 126, and cross at PDF 189, more cross on the 15th at PDF 225, redirect at PDF 261, and recross at PDF 269.

If you want to see and hear it, go to the
video archives.

TBV

tbv@trustbut.com said...

Timm,

The official results in the LDP were manually adjusted, for what Brenna called "QA purposes".

There was a "reprocessing" of data in April that ran the same data sets on the other machine in various modes, including "full auto"

Different people draw different conclusions from the results of the processing and reprocessing -- not surprisingly in line with their mind-sets going in.

TBV

Michael said...

TBV,

A great exercise, but how do we get this out to the public.

And does it prove that Floyd is innocent or falsely accused? If so, who'd going to care but us?

Thanks,

Mike

BustinBilly said...

Hatefrench - Dr. Brenna's comments on the blank reprocessing are on page 179 of the official transcript.

TBV - All eight of the 5alpha androstandiol-5beta pregnandiol reprocessed differences exceeded
-5.5 per mil. The reprocessing was requested by Floyd's side.

Four of the other seven samples had differences in excess of -4.6 per mil. The integration on those samples was done automatically by Masslynx.

I believed Floyd until I studied the evidence last spring.