Tuesday, July 30, 2013

An interview with David Balding

Part 36 in the Knox/Sollecito case

Professor David Balding recently published an analysis of the bra clasp DNA.  It may be helpful to explain some terms found in this article.  John Butler (Fundamentals of Forensic DNA Typing) defines the liklihood ratio (LR) as “The ratio of the probabilities of the same event under different hypotheses, and he explains that the prosecution’s hypothesis is usually the numerator, and the defense’s hypothesis is usually the denominator.  A ban is a unit of expressing the weight of evidence (WoE).  This scale is logarithmic; a likelihood ratio of three bans is equal to one thousand.  Some months ago Dr. Balding was kind enough to answer some of my questions about this work.

Does Raffaele Sollecito¹s DNA fall into the category of low template DNA,
and if so, should two separate amplifications have been run?

There's no strict definition of "low-template" but broadly yes the peaks associated with Sollecito are low (but not those associated with Kercher, they are high).  Conti-Vecchiotti discuss a threshold of 50 rfu below which a peak should not be relied on; in the UK, that threshold was used in the past but nowadays as techniques have improved the threshold is often lower, 25 or 30.  However that doesn't matter here as all the peaks associated with Sollecito are well above 50: there is a 65, a 70 and a 98, all the 26 other peaks are above 100.  So it is not extremely low template - many low-template cases are successfully prosecuted in the UK even when some peaks fall below the threshold and so are discounted.  In this case all the peaks associated with Sollecito seem clear and distinct  so I think there can be no concern about the quality of the result as far as it concerns him or Kercher.

Replication is generally a good thing and is nowadays done in most cases in my experience, but not all - one problem is that replication splits the sample and so can increase the chance of not getting a usable result.  But although replication is desirable it is not essential.  (In a sense there already is replication, because each of the 15 loci is an independent test.)  This is all a matter of weight of evidence, which Conti-Vecchiotti paid no attention to: if you measure the weight of evidence properly, that accounts for the extra assurance that comes from replication and gives a stronger result (or conversely gives a weaker result if there is not replication).  But because Sollecito is fully represented in the stain at 15 loci (we still only use 10 in the UK, so 15 is a lot), the evidence against him is strong even allowing for the additional uncertainty due to non-replication.

Are there contributors other than Raffaele Sollecito and Meredith Kercher to the autosomal profiles?  If so, how does the presence of this additional DNA affect the bra clasp as evidence?

Yes, Conti-Vecchiotti identified a further 12 above-threshold peaks at alleles that could not have come from Sollecito or Kercher.  They correctly criticised the scientific police for ignoring these: many do appear to be stutter peaks which are usually ignored, but 4 are not and definitely indicate DNA from another individual.  The extra peaks are all low, so the extra individuals contributed very little DNA.  That kind of extraneous DNA is routine in low-template work: our environment is covered with DNA from breath and touch, including a lot of fragmentary DNA from degraded cells that can show up in low-template analyses.  There is virtually no crime sample that doesn't have some environmental DNA on it, from individuals not directly involved in the crime.  This does create additional uncertainty in the analysis because of the extra ambiguity about the true profile of the contributor of interest, but as long as it is correctly allowed for in the analysis there is no problem - it is completely routine.

Are there contributors to the Y-STR profile other than Raffaele Sollecito?  If so, how does the presence of this DNA affect our interpretation of the bra clasp as evidence?

I haven't looked closely at the Y evidence - there seems no need for it because the autosomal evidence is overwhelming for the presence of DNA from Sollecito.  But from a look at Conti-Vecchiotti, it seems to back up the conclusion from the autosomal profiles: Sollecito's alleles are all represented and these generate the highest peaks, but there are some low peaks not attributable to him; so at least one of the additional contributors of low-level DNA to the sample was male.

The bra clasp was collected about 47 days after the murder, and it was found in a different location from where it was initially observed.  In the interim many people entered the cottage and items from her room were removed.  Are these concerns sufficient for the clasp to be excluded as

The only worry would be if somehow DNA from Sollecito was brought into the room and deposited on item 165B.  I don't know enough about what happened to say if that was likely but I'd guess that people walking in and out of the room etc would be unlikely to do that.

The clasp was collected with gloves that were not clean, not with disposable tweezers (videos here and here).  The glove was handled by more than one person.  Are these concerns sufficient for the clasp to be excluded as evidence? If not, should the clasp be given less weight as evidence because of them?

Same comment - the only concern is if any of this could have transferred DNA from Sollecito onto item 165B.

Would you care to comment on the storage of the clasp after the forensic police tested it?

I know nothing about it.

Did you analyze the electronic data files?  Did you examine the laboratory¹s own protocols and machine logs?

I have only seen the epgs for the autosomal DNA profiles of 165B.  There is an unclear version of them in the Conti-Vecchiotti report, but Prof Vecchiotti kindly provided my with a clean set.

Did you examine the negative controls?


Tuesday, July 9, 2013

The sensitivities of presumptive and confirmatory blood tests

Part 35 in the Knox/Sollecito case


Forensic tests for blood have been the subject of four previous entries, one of which discussed the pseudoperoxidase activity of hemoglobin, one of which discussed the difference between presumptive and confirmatory testing, one of which covered the difference between mixed DNA versus mixed blood, and one of which covered some case histories and which also treated the use of tetramethylbenzidine (TMB) in forensic testing.   Confirmatory tests for blood can be performed by testing for the presence of a biomolecule that is unique to blood:  hemoglobin (the oxygen-carrying protein found within red blood cells), immunoglobulin G (an antibody found in plasma), or glycophorin A, (a membrane protein found on the surface of red blood cells). 

Executive Summary

Positive presumptive blood tests indicate the possibility of blood, but only confirmatory tests allow for the conclusion that blood is present.  Modern confirmatory testing of blood is extremely sensitive, yet was apparently not used on the luminol-positive areas in the present case.  The reported sensitivities of presumptive tests for blood have varied widely. Among the reasons for the differences are differences in concentrations of the chemical reagents, in the times of reaction, whether or not the samples were dried, and whether the reagents were tested under laboratory or field conditions. Luminol is very sensitive, but the intensity of the chemiluminescence decreases with increasing dilution.  Luminol is at best only slightly more sensitive than tetramethylbenzidine (TMB), and any difference is dependent on the particulars of how each test is performed.  The luminol-positive areas in this cases tested negative by TMB; moreover, there was no reported testing of these areas by confirmatory experiments.  Therefore, the luminol-positive substance or substances is (are) more likely to be something other than blood.

A brief overview of confirmatory testing of blood

The National Forensic Science Technology Center wrote, "The line between screening and identification is not always clear. For example, while examining the clothing of a suspect, a forensic biologist might visually locate a brown stain that presumptively tested positive for blood and was then DNA typed. The DNA type is found to match the victim. Knowing that the loci tested are higher primate specific, what conclusions can be drawn?  The only unqualified conclusion that can be offered is that the stain contains DNA that matches the victim. It has not been proven to be blood.  If asked ‘Could the results have arisen because the material tested was the blood of the victim?’ then an answer of ‘Yes’ is justified. However, it would be wrong to report that the material was human blood with a DNA type that matched the victim. The material was not subjected to confirmatory testing for blood or proven to be human in origin."  It must always be borne in mind that the burden of proving the existence of blood falls on the prosecution; it is not the job of the defense to prove that a substance that gives a positive result in a presumptive test is not blood.

Modern confirmatory testing for blood owes its specificity to the use of antibodies, proteins which recognize (bind to) certain molecules (antigens) and which generally do not recognize even closely related molecules.  Modern confirmatory testing owes its sensitivity (see below) to the use of enzymes that are covalently linked (conjugated) to the antibodies.  Such tests are called enzyme-linked immunosorbent assays or ELISAs.  Enzymes are catalysts; therefore, for each antibodt/antigen complex that is formed, perhaps thousands of substrate molecules are converted into products, which may be easily detected because they are often colored.  ELISAs come in a variety of forms that are beyond the scope of the present article. 

The general lack of confirmatory testing in the Knox/Sollecito case

The lack of confirmatory testing greatly weakens greatly the prosecution's conjecture that the luminol-positive areas in the Knox/Sollecito case were the result of blood. A true confirmatory test was apparently never performed on these areas; therefore, one cannot conclude that luminol-positive material was blood.  These areas also returned negative results in the TMB tests.  Yet the forensic police did use confirmatory testing on Rep. 199, which came from Filomena’s room, indicating that they had the ability to perform such tests.  With respect to the luminol-positive, DNA-negative areas in this case, Drs.Virkler and Lednev said, “The prosecution should have used much more convincing evidence to prove the presence of blood.”  Indeed.  Confirmatory tests have become rapid and sensitive; therefore, it is difficult to see why one would not perform them.

The sensitivities of two confirmatory tests for blood

The calculated and reported values of the sensitivies (below) might be too optimistic by a factor of ten, and yet still they would indicate that confirmatory tests for blood are extremely sensitive.  One possible problem is that eventually proteins such as hemoglobin denature, meaning that they lose their biological activity (denaturation implies a change in the three-dimensional shape of a protein or that they undergo chemical alteration).  Denaturation could keep a protein from reacting in confirmatory experiments. That is why the several-month delay in testing Rep. 199 using a confirmatory test was so unfortunate.

Whole blood is about 55% serum and 45% red blood cells (RBCs), although there is some variation in this ratio among individuals. Hemoglobin (Hb) is found in RBCs and immunoglobulin G (IgG) is found in serum.  My calculations of the sensitivities of confirmatory blood tests attempt to normalize to whole blood.

Robert Kerber wrote, "The concentration of hemoglobin molecules in red blood cells is so high (340 mg/mL, 2.3 mM) that they almost could be said to be on the verge of crystallization." (J. Chem. Education Vol. 84 No. 9 September 2007, p. 1541). Therefore, the concentration of hemoglobin is 150 mg/mL in whole blood. 150 mg/mL divided by 2.2 x 10-5 mg/mL = 7 x 106. This calculated dilution factor for detection of hemoglobin is close to the reported values for the HemaTrace test.  One article from the Michigan State Police listed its sensitivity as 0.05 µg/mL, and the authors gave its maximum dilution factor of blood that would still allow for its detection as 1:16,777,216. Another study provides 0.07 µg/mL as the limit of detection.

Williams and coworkers (Forensic Science International Vol. 190, 2009, pp. 91–97) described a sandwich ELISA protocol for immunoglobulin G that have a detection limit of at least 0.1 µg/mL, possibly higher.  ELISA stands for enzyme-linked immunosorbent assay, and it is a type of assay that uses antibodies to bring about specificity.   In the textbook Principles of Biochemistry, Mammalian Biochemistry, seventh edition (McGraw Hill, 1983), Smith et al. indicated that the average concentration of IgG is 12 mg/mL in plasma, therefore its concentration is 6.6 mg/mL in whole blood.  Along with the detection limit of 0.1 µg/mL, this concentration suggests that the test for IgG should detect blood diluted up to a factor of 6.6 x 104, similar to the value of 100,000 reported by Williams and coworkers.

Reports on the sensitivities of presumptive blood tests

The following list is an attempt to survey the forensic literature for information on the abilities of presumptive tests, especially luminol and TMB, to detect diluted blood.  There are wide ranges for the reported sensitivities of each.  A previous blog entry rebutted the suggestion that the reason for the lack of positive TMB results was the supposed greater sensitivity of luminol vs. TMB.  One important conclusion of that entry is that it is commonplace in forensics to follow a fluorescent test such as luminol with a colorimetric test such as TMB.  From p. 258 of the English translation of the Massei report, “She [defense expert witness Sarah Gino] added that, in her own experience, analyses performed with TMB on traces revealed by Luminol give about even results: 50% negative, 50% positive, [276]

L. Garofano, M. Pizzamiglio, A. Marino, A. Brighenti, F. Romani.  "A comparative study of the sensitivity and specifity of luminol and fluorescein on diluted and aged bloodstains and subsequent STRs typing," Int. Congress Ser. 1288 (2006), pp. 657–659.

Kastle-Meyer: 1 part in 10,000
Leucocrystal violet: 1 part in 10,000

Emma Johnston; Carole E. Ames; Kathryn E. Dagnall; John Foster; and Barbara E. Daniel.
J Forensic Sci, May 2008, Vol. 53, No. 3, “Comparison of Presumptive Blood Test Kits Including Hexagon OBTI”

Hemastix 1 part in 50,000 [note: Hemastix is based on the TMB test]
Kastle Meyer 1 part in 10,000

Filippo Barni, Simon W. Lewis, Andrea Berti, Gordon M. Miskelly, Giampietro Lago. Talanta 72 (2007), pp. 896–913, “Forensic application of the luminol reaction as a presumptive test for latent blood detection”

“Luminol can be used to detect the presence of minor, unnoticed or hidden bloodstains diluted down to a level of 1:106 (1 µL of blood in 1 L of solution) [18,63,72].”

Joanne L. Webb, Jonathan I. Creamer, and Terence I. Quickenden.  Luminescence 2006; 21: pp. 214–220, DOI: 10.1002/bio.908, “A comparison of the presumptive luminol test for blood with four non-chemiluminescent forensic techniques”

Hemastix 1:1,000,000 solution
Luminol 1:5,000,000 solution
Both values are said to be consistent with the literature.
“What is most apparent is that the literature available on presumptive blood detection techniques is somewhat variable. There is a great diversity of experimental conditions, which makes comparison between reagents tested by different authors difficult.”

Cox, M. “A Study of the Sensitivity and Specificity of Four Presumptive Tests for Blood,” Journal of Forensic Sciences, JFSCA, Vol. 36, No. 5, Sept. 1991, pp. 1503-1511.

TMB gives a fast positive result at 1/10,000 dilution; a slower reaction is seen up to 1 to 1,000,000 dilution.

Shanan S. Tobe; Nigel Watson; and Niamh Nic Dae´id.
J Forensic Sci, January 2007, Vol. 52, No. 1, doi:10.1111/j.1556-4029.2006.00324.x, “Evaluation of Six Presumptive Tests for Blood, Their Specificity, Sensitivity, and Effect on High Molecular-Weight DNA”

“In the past 50 years, there have been many tests conducted on the sensitivity of presumptive blood tests (4,6–15). The findings of these studies are in great contradiction with each other. Sensitivities for luminol range from 1:200 (11) to 1:100,000,000 (6); from 1:200 (11) to 1:100,000 for leuchomalachite green (LMG) (8); and from 1:2,000 (12,13) to 1:10,000,000 for phenolphthalein (9). The various differences in the sensitivities reported by different researchers of presumptive blood tests are probably caused by differences in reagent concentrations, methods of preparation of samples, reagents and results, and in the type of material containing the blood (4).”

From Table 2
Luminol 1 part in 100,000
Hemastix 1 part in 10,000 within 1 minute; 1 part in 100,000 within 2 minutes

“The luminol reagent reacted instantly, with both the 1:10,000 and 1:100,000 dilution factors producing a blue luminescence. The luminescence lasted for close to a minute. However, both dilution factors were much less intense than the positive control of whole blood. The reaction with the 1:100,000 dilution factor was extremely faint.”

“The Bluestar reagent reacted instantly with the 1:10,000 with a blue luminescent glow but faded within a few seconds. The 1:100,000 dilution showed slight reactivity, with five of the 25 samples showing a very faint positive, which faded in a few seconds. However, both dilution factors were much less intense than the positive control of whole blood.”

Anders Nilsson.  The Swedish National Laboratory of Forensic Science (SKL), Linköping 2006, “The forensic luminol test for blood: unwanted interference and the effect on subsequent analysis”

“Under laboratorial conditions CL [chemiluminescence] was detected from luminol treated stains of the used hemoglobin solution (corresponding to blood) diluted up to 5·106 times. A comparably high sensitivity of the luminol test has been reported in other studies [22].  However the sensitivity is probably not as great under the conditions found at a crime scene and here, depending on several factors, perhaps one may “only” see blood diluted to about 1:10000 [14].”

D. D. Garner; K. M. Cano; R. S. Peimer; and T. E. Yeshion.  Journal of Forensic Sciences, Vol. 21, No. 4, 1976, “An Evaluation of Tetramethylbenzidine as a Presumptive Test for Blood”

“As shown in Table 1, both reagents at the 0.05 M concentration will detect one part
blood in 10 000 parts isotonic saline. Doubling the concentration of the reagent results in
a tenfold increase in sensitivity for both TMB and benzidine. The lowest level of detection of blood by both chemicals was 1 ppm.”

LJ Blum, P Esperanca, S Rocquefelte.  Can. Soc. Forensic Sci. J. Vol. 39. No 3 (2006) pp. 81–100, “A new high-performance reagent and procedure for latent bloodstain detection based on luminol chemiluminescence”

Dilution Light intensity a.u.
1:5                           314,600
1:100                         89,720
1:1000                         6,725
1:10000                        1730
“As expected, the lower the dilution factor, the lower the light intensity (Fig. 7).”

General reference on presumptive and confirmatory tests for blood, semen, and saliva
Kelly Virkler, Igor K. Lednev.  Forensic Science International 188 (2009), pp. 1–17,  “Analysis of body fluids for forensic purposes: From laboratory testing to non-destructive rapid confirmatory identification at a crime scene”