Spoor #10 scrutinized

The Dutch Forensic Institute (NFI) claimed trace #10 to match only with the DNA-profile of Louwes. During the petition for review of the sentence in 2006 something changed. The original electropherogram of trace #10 among others was submitted to the defence counseler Prof. Mr. Knoops at last, revealing several additional peaks matching the profile of the victim. The NFI analist had signalled those additions using a special am-label, meaning added manually. In two cases they coincided with peaks present in the profile of the victim as homozygotes, so appearing in her profile as peaks of doubled height: D3S1358 and D8S1179.

A bit to much of a coincidence.

This reconstruction of the electropherogram of trace #10 displaying 11 loci to be discussed below.

Zooming in

Below all positions ('loci') in the reference buccal swab profile of Louwes are compared to the profile in trace #10.

A major complication when trying to point out a additional ontribution of a second person with a known profile (POI = person of interest) consist of the circumstance that a great number of positions is already taken by profile of the first POI. To kind of occupations are possible. First of all, a peak can be a true allele, just an ordinary peak in the profile, Normally there are two of them. If they overlay the contribution of the second POI, then this overlay as hardly or not visible, even more if the contribution of the second POI is small. Yet there is an indication. Most of the time, two peaks occupy a locus, their heights are rather evenly balanced. Rather because there are potential causes for imbalance. This imbalance can be put in a figure, the ratio in peak heights. A distinctive change in balance can be an indication of an addition from another source. Of cause only if the involved peak height increases. This can be shown in five occasions. One occasion can be coincidental. Even twice. But five times?
The second kind of occupation we meet is caused by the amplification process in use during PCR. A lot of fragments after amplification are to short compared with the parent fragments. In the electropherograms they appear as co-called backward stutters. They are shorter then the parent peaks, typically about 5 to 15 % in height. Nowadays it is custom to compare the height with validation data to decide if the peak is a stutter peak or perhaps a combination of several peaks (stutter filtering). In four cases the stutter peak of the first POI is potentially increased with a peak of the second POI. Also check appendix 1.
In about four cases - not all equally distinctive - we observe peaks in new positions, possibly true alleles of the second POI. Twice, this is in the position of a so-called forward stutter. This type of stutter is very low in height most of the time (1 - 2%) so the addtional height is easily recognized. Again check appendix 1.

Also, in about thirteen positions we observe an irregularity in the profile of trace #10 whem comparing the profile with the reference profile of Louwes, in all cases indicating a potential effect of the presence of DNA of Ms. Wittenberg. Is that enough? Of course we will have to take into account the intensity of the presence of her profile, 10 to 20 times as weak as the profile of Louwes. This is logical inrespect to the size and colour intensity of the blood spot involved occupying only een small part - about 1/20 - of the cutting of trace #10. Check 6.3.2: Vlek #10 uitgerekend.

A profile of comparable intensity of the victim is presented elsewhere in the file, being trace #6. The reference profile of W consists of 18 peaks. In profile #6, there only 12 or 13 peaks, the other ones being drop-outs it appears, so the number of 13 potential peaks should be enough for identification.

By the way, when the NFI found iregularities in the profile of trace #10 (the am-markers), one would expect the NFI to exercise replicates, preferably starting with the original extract (400 μLwas available, so a lot more than in case of the extracts of later samples extracted with QUAamp) as well with the PCR-product, which is abundant. You never know... Or, were replicates measured and are we only confronted with the only one fitting best with other results? This seems odd, but otherwise: the parameters used in the analysis of trace #10 differ from the parameters advised by the manufacturer in two instances. This happened several times in this case. But normally it was the result of measuring and presenting replicates. So did replicates exist and how did they look??

Another: another institute (FLDO Leiden) was asked to give a second opinion on basis of the traces #1 - #9. When the samples were sent to Leiden, the NFI had completed the analysis of the samples 10 - 20. Why did not FLDO receive this samples too???

To extend the latter issue: in 2006 the NFI analyzed trace #42, sampled near trace #10, again displaying a tiny, even tinier blood spot, and characterized it again as an unmixed profile of Louwes. This sample indeed found his way to FLDO this time and was subsequently recognized as a mixed profile of Lowes and Ms. Wittenberg. Other glasses? Other stance?

Locus and references (zie legenda below)	referenceLouwes	trace #10	comparisons
D3 l 14 14 L 15 15 W 16 16			Trace #10 reveils a true allele of W16W16 on the forward stutterposite 16 of LL15. Forward stutters most of the time are extremely low, too low to explain the added peak. Correctly the NFI marked the peak but omitted this from the subsequent report.
VWA l 15 (16) L 16 17 W 17 19			Imbalance. Locus VWA17 in trace #10 increased proportionally compared to the reference profile due to W17. The balance is overtrown. W19 is invisible (drop-out).
Locus en referentiesporen	reference Louwes	trace #10	comparisons
D16 l 11 (12) L 12 13 W 11 12			Here both contributions W11 and W12 are visible. W11 increased the backward stutter l11 above the threshold. At the same time W12 overthrew the imbalance L12-L13.
D2 l 18 23 L 19 24 W 20 23			W20 is invisible (drop-out), W23 might contrbute to l23, which increased in height and rose above the stutter threshold. Note, the current imbalance was not overthrown. There was no reason to expect so.
Locus and references	reference Louwes	trace #10	comparisons
AmelXY L X Y W X X			Here the imbalance is overthrown significantly possibly because of the added contribution of XX from the DNA of W.
D8 l 7 12 L 8 13 W 14 14			The true allel W14W14 is visible and marked in the position of the forward stutter l14 van L13, exceeding the threshold by far.
Locus and references	reference Louwes	trace #10	comparisons
D21 l 26 30 L 27 31 W 31 31.2			The imbalance of L27 and L31 is almost levelled, possible under the influence of W31. In addition some irregularities remain unexplained (24 and 29.2). W 31.2 appears invisible (drop-out).
D18 l 12 22 L 13 23 W 15 17			W15 seems to be visible. Furthermore a small peak in both profiles (18) remains unexplained (contamination?).
Locus and references	reference Louwes	trace #10	comparisons
D19 l 12 14 L 13 15 W 14 14.2			Contributions of W might be visible at 14 and 14.2 At 14 the heigt of the stutter increased significantly (from 10 % to 13% exceeding the stutter threshold l14). In both profiles tiny irregularities at 14.2.
THO1 l 6 8.3 L 7 9.3 W 7 7			A marginal overthrow in imbalance is visible in agreement with a contribution of W7 W7. Stutters are not relevant and also very low.
Locus and references	reference Louwes	trace #10	comparisons
FGA l 22 22 L 24 24 W 22 24			The contribution W22 is potentially the cause of the increase of stutter l22 22, now exceeding the stutter threshold. Possibly there is a peak at 24 (W24), but this is very unclear.

Legenda
*	peak with low RFU in trace#10( with a similar marker in the reference profile).
l	to point at stutter peaks of Louwes. Because the RFU-values of the stutters in the profile of the victim are expected to be extremely low, they are ignored.
L	true alleles of Louwes.
W	true alleles of the victim Wittenberg.
xx	peak visible in trace 10 but lacking in the reference of Louwes.

Comparing trace #10 to the reference profile of Louwes reveals a number of differences, which can be used to construct a new electropherogram. In five positions no trace of the victim is visible, just in several other profiles of the victim in the file of the Deventer Murder Case.
The presence of traces at other positions justify the position that the presence of DNA of the victim in trace #10 cannot be excluded.
So trace #10 is a mixed profile and therefor it is possible that the blood spot visible in trace #10 originates from the victim.
This is in agreement with the evaluation of the peak heights involved as demonstrated in the paragraph calculations.

The electropherogram of trace #06 exhibits a comparable set of peaks of comparable heights. This trace was considered a partial profile of the victim on base of a frequency of less than 1 out of billion (NFI report 5 december 2003). In 11 or 12 positions (out of 17) it matches her reference profile, so largely in agreement with the reconstructed profile aforementioned.

An overview

Another approach is the following. We might compare all peaks and tiny peaks in trace #10 to the reference profile of Louwes and calculate the measure of correlation. Of cource, the agreement will be strong, trace #10 consist in both hypothesis of the defence and the procecutor largely or completely of DNA of Louwes:

Horizontal: peak heights in the mean reference profile of Louwes (there are two replicates). Vertical: matching peak heights in the profile of trace #10. Correlation according the R-squared value is very high (mximum is 1). When fossusing on the lowest peaks - see inset picture - the R-squared value is very low in contrast.

Next, a hypothetical profile was calculated as follows: to the mean reference profile of Louwes add 1/6 part of the reference profile of the victim. The reference profile of the victim - mesured in peak heights - contained only 1/3 of DNA compared to the reference profile of Louwens, so the share of W in the final result amounted to 1/20, as was the estimated the amount of blood in the sample, based on size and intensity of the colour of the blood spot in it. Check Vlek #10 uitgerekend for details. Following the same procedures as before, we see a rise in correlation, so demonstrating this approach (trace #10 is a mixture) to be superior to the previous one.

Horizontal: peak heights in a mixed profile of Louwes with an addition of 5% profile of the victim. Vertical again peak heights in trace #10. The R-squares values are higher, notably in the lower quadrant.

De likelihood ratio

A method exists to compare the proof of the prosecutor and the defence of the defence counseler. In this method two hypotheses are compared: Hp of the prosecutor and Hd of the defence. The ratio Hp/Hd is thought to balance the weight of proof. Because eveyone is presumed innocent a priori, a LR (likelihood ratio) of Hp/Hd needs to be rather high.

Of course there rises a question, how to attach values to Hp and Hd? In the case of DNA-proof there are calculations supported by digital techniques. And so we are at the mercy of a black box. But the princinples within are clear: not only the occurence op peaks are relevant, but a lot of details too like the height of stutter peaks and the possible combination with peaks of a second DNA trace, the influence on peak balances, just als aforementioned. The total of peaks is correlated to expectations leading to likelihoods for both sides.

Lets take an exemple. Above, about ten indications are presented to support the hypothesis of a mixed profile W+L. Let presume and exaggerate, each one of is in favour of Hp with a factor of 90%. In each case, the prosecutor has to be teh winner, else, the sample is a mixture. This means, from the prosecutors view, that probablity the overall hypothesis is true amounts to (0,9)¹⁰ = 0,35 . So there is a 65% probality, that the hypothesis is false!

And Hd? That is a different case: just one strong indication is enough to establish the sample to be a mixture. In this case there two of those strong indication, the peaks on the forward stutters are far to high to be a coincidence (calculations show likelihood ratios of 3.5 and 40 in favour of Hd!).

Also check the chapter covering DNA.

Appendix 1: Stutter heights

Backward stutters

Four peaks - matching with the profile of he victim -connected (combined with) to backward stutters exhibit peak heights surpassing easely peak heights measured during validation of yhe used SGMPlus system as shown in the manual of the kit as well in comparison with the medians as with their maxima. In comparison with the mean values, this peak heights in trace 10 are 4 to 9% higher, in agreement with the estimate of 5% of DNA of the victim in trace #10.

Forward stutters


The forward stutter in D3 has a maximum of 1% hoog (open circles), or RFU 40 (solid circles) where a match with the victim is suspected (red vertical line). In trace#10 we see een peak height of 94 RFU or 2,2%, so double in height. This peak matches with the profile of the victim, with a homozygotic peak notably.	This forward stutter in D8 has a maximum of 2% or 40 RFU. In trace #10 we find a peak height of RFU 115, or 4,4%, again twice in height. Again matching to a homozygotic peak in the victim's profile.

Both on the same spot

Peak 14 in locus D19 is special, being a forward stutter of L13 and a backward stutter van L15. De backward stutter should be on average 8% of the parent with RFU 1309 amounting to 105 RFU. De forward stutter contributes about 30 RFU. In reallty the peak measures 173 RFU, so 40 RFU higher.

Appendix 2: Peak Height Ratio (PHR)

The amount of shift is expressed in proportion to the standard deviation measured during the validation. Based on a normal distribution, a shift of 2 x SD shifts a value from the centre of the distribution to a zone with 2,5% probability (it being coincidental). Because the definition of PHR is rather peculiar, the distribution is possibly not normal, but in some cases , the shift is far more than 2 x SD and might be considered significant nevertheless.

Most loci are filled with two prominent peaks of Louwes (heterozygous and XY-amel). Exceptions ('homozygoot' ) are indicated. Originally both peaks ought to be of the same height. In reality, there is always a difference; the column PHR (Peak Height Ratio) shows the most probable imbalance based on a number of validation experiments (50 - 100) by the manufacturer of the SGMPlus kit, which was in use at the NFI. There is no mention of the fact which peak was higher, the left or the right one. Using data of the NFI, the PHR's of the mean reference sample of Louwes was calculated. The value of D18S51 deviates strongly and has to be left out of the picture (but observe the huge interval between the two peaks). In this case without consequences, an extra imbalance is not to be expected, because there is no overlap between L and W at this locus. Where an imbalance is to be expected, the data are marked yellow. The column 'spoor #10' shows the new PHR, complying to the original sequence of the PHR. The suspected addition of DNA-STR's from the profile of the victim led almost everywhere to a reversion of polarity; the smallest peak of the couple almost everywhere switched to be the largest. By chance the addition is almost every instance combined with the smallest peak of the couple so the change afterwards gets very clear.