Still milling. Don't know if we're getting rebuttal, or going to arguments.
YOUNG: put Davis' declaration as ex 131.
DIRECT by MR YOUNG
q: we heard Dr. Aug say the only way he could identify the internal sample by "divine intervention"
q: did you have any problem identifying the internal standards?
a: no, not at all.
q: how did you know where it was.
a: Have to take you through some of the documentation.
Young hands out out documents
q: By the way, the only symbolism in my tie is that my daughter picked it out.
a: USADA 182, first mix cal acetate. Four peaks because we built the standard. We see it as the first. If we flip to USADA 181, and we look at values for time (866 1229 1302 1473). So we know standard is at 866. USADA 18, mix cal acetate done after, with whatever intervening time. Then on USADA 183 (?) retention is 866 1228 1301, 1471) pretty close across the board. This gives us some confidence the peak at 866 is the internal standard in the intervening runs.
a: Look at samples, USADA 158, the CG, and USADA 157 is the sample time (867, 1473, 1763, 1826). From my experience, 876 is the closest peak, and that is within error, so I choose that one as the internal standard.
a: will go through the rest, because there aren't many; USADA 164, USADA 163, (868 1231 1247); so that is it; next is f3 from blank - more complex - USADA 170, USADA 169 (798, 867, 1244, 1306, 1336, 1375, 1651), so pick that one.
BRUNET: by what standard would you say it's off?
a: wait. I'll address that.
a: USADA 168, cg, f1; USADA 160 (778, 867, 881, 1236, 1478, 1764, 1824); the consistency amoung these CG's say that were ok picking 867, will explain more later. USADA 167, USADA 166 (866, 1229, 1254, 1288), and that's the internal standard. USADA 173, USADA 172 f3 sample (795, 827, 867, 1304, 1337, 1652); other peaks are far away.
a: B samples USADA 361, a mix-cal-acetate, USADA 360 (870, 1241, 1316, 1491). That's bigger, but it's a very different set of conditions with the machine.
YOUNG: if time was within a couple of weeks, make a difference?
a: USADA 363, mix cal at end (870, 1241, 1316 1490); now look at samples USADA 333, USADA 335 (869, 1489, 1843); USADA 341 (869, 1241, 1267), USADA 347 (799, 872, 881, 1260, 1323, 1353, 1394, 1672), will explain ambiguity; USADA 338 (777, 870, 1490), USADA 344 (870, 1241, 1266, 1301),
a: USADA 349 shows CG of a congested sample, similar to one used as an example. USADA 350 (797, 839, 871, 880, 909, 1318, 1352, 1671).
[ Landis is drilling holes with a glare ]
Pulling up a slide Brenna made, showing internal standard times.
a: A sample mean 866.6, +/ .07%, B 870.5 +-/.13%; so 880 is not the internal standard.
q: are these within 1%?
a: yes, .07%
q: +/- .2 minute.
q: do you expect GCMS retention to be the same as the IRMS instrument?
a: no; a diagram would help;
[ uses Davis' block diagram; he's going to draw ]
a: the gcms is a faster detector (nearly instant); in the irms the sample gets burned, dried, and sent through plumbing before getting to irms. So there is latency between the GC part and the IRMS part. They might be proportional, but it usually doesn't work that way, because the speed through GC depends on the molecule, but through the IRMS it's all CO2 at the same speed. This has implications for calculating retention and relative retention times.
Young: make exhibit.
q: in your lab, would you expect gc and irms times to correspond?
a: No. we run every day, and match peaks every day.
q: relative retention?
a: no, for reasons outlined.
q: believe any linearity problem during S17 processing?
q: why not?
a: we've been provided linearity tests.
q: LNDD0312, despite Dr. Davis saying not saying received:
OBJECTION: Mischaracterize, said he had not seen runs, not data.
q: have you seen results form Jun, Jul, September?
q: start at 314, and explain.
a: I'll repeat that we want to know ratio is accurate at small sizes and large sizes. Here we inject reference gas at different volumes in pulses.
q: is the purpose to show the instrument is giving the same result at these heights?
a: yes; to show yardstick is accurate at 6 inches and 6 feet.
q: so what are we looking for?
a: in the data that follows, we want to see those ratios are nearly the same. This data used in conjunction with samples, you'd want to bracket the data with the sample values.
q: is that true in S17.
a: yes, it was; there's discussion it's not the case. I want to look at signal intensities, which is usual. I can't speak to an error made elsewhere. When I compare the intensity here with those in the samples, they are above these lower levels.
[ that's an interesting hedge there. Crap in shoulders and background that is non-linear of unknown composition can throw things off. ]
q: if there was a linearity problem... do you have any concern there is a linearity problem?
q: if there was a linearity problem, would the problem be reflected in peaks that have very different heights?
a: the lowest peak in the lin was down here; all the 5a's are way above that. You'd expect a linearity problem to be reflected when one compares peaks of quite different sizes. While i won't say this size difference is enough to cause that problem, these two are very close in intensity
[ misses the point of background co-elution of differnt substance or values, and integration errors, as demonstarted by the -70 at 5% example ]
q: have you heard any challenge to the mix-cals in 8 days?
OBJECTION: improper direct.
a: I don't recall hearing any. I did hear a number of things that were consistent with my opinions.
q: if they are working, does that mean the instrument is working?
a: yes, it means it is working well for resolved peaks
[ Another dodge -- Herr Doktor's point was whether the instrument worked in complicated matrices with non-well resolved peaks ]
q: if there was a pressure problem, would you have the consistent mix-cal we have here?
a: it would be reflected in a diagnostic. The thinking is a little bit odd to me. I'd look at the mix-cal and mix-cal-acetate, and if the results were not good, then I'd look at pressure or something else.
q: explain the concept of of overlapping peaks, and which contributes to the other. Your paper was quoted, and I want to get it right.
[ draws picture, two -28's, resolved, then overlapped ]
a: our study started with real data, not modelling only. It showed a direction opposite of what was said, we got -29 -27, and that's what I'd expect here, if it were an artifact, which I don't believe it is.
[ landis smiles, then breathes deeply. ]
q: let me take you to just two chromatograms. first is A sample diol, second the B. In reviewing these and the corresponding GC chromatograms, and the ion traces, do you see any problems with interference or other problems that would cause these results to be unreliable.
a: I do not.
q: Dr. Davis showed an example of moving the baseline, and sometimes seeing significantly different results? Is that the way you saw them do it in LNDD?
a: no, it is not.
q: explain the differences in what Davis did and the LNDD techs?
a: first, I say the techs adjust the data, which I call a QC step, and will do here; I observed closely and they first interested in peak start stop, but in verifying the background was properly modelled. The x's you saw deleted and moved, called background points, are better described as anchor points. The s/w uses those to calculate background. Older (and newer) algorithms sometimes choose anchor points that are on peaks. So part of their QC is to be sure the anchors are on background, not peaks. They would quite often move anchor points to ensure to their satistaction that the calculated background corresponded with the trace on the screen. And they took care ensuring that. When I asked them later, that is what they said they were doing.
q: that in the SOP?
q: When he did that this morning, was he setting on peaks?
a: yes, and the results were not in the least surprising; that is what the SOP is intended to avoid, and why it is a QC step.
q: peak start stop. When Dr. Davis was doing this, he appeared to be looking at the peak chromatogram. What did the LNDD techs do?
a: they were looking for something specific, a peak start should have ion/current isotope ratio corresponding to background. Also the swoosh. They'd call up the box, of ion intensity and 2/1. They'd move the cursor to a region that clearly had no peak. If memory serves, at least for the A's the number was 1.174, would advance to see change, back off, and set at a background point, and again on the other side of the peak. I was looking carefully, and they were doing the same thing every time.
[ Duckstrap notes: Brenna appears to be comparing the GCMS chromatograms to alternative GCMS CGs. These are similar, as expected because the chromatography conditions are similar. What is not similar are GCMS CGs that should lead to peak identification, and the IRMS CGs that produce the CIR values. These chromatography conditions are not similar to the GCMS. These are the ones that need to match, and they do not, by a mile. ]
[ Also didn't address magnet, or storage of parameter info ]
SUH: a few questions for Davis after cross.
Barnett: We'll have no more witnesses; Landis will not ask Botre anything;
We'll get Brenna cross, then a couple for Davis, then arguments this afternoon.