NEW YORK (GenomeWeb) – Researchers at the University of Tuebingen are applying mass spec-based proteomics to bullet analysis in forensic investigations.
In a paper published last month in the Journal of Proteome Research, the researchers demonstrated that they were able to generate organ-specific proteomic profiles for material found on bullets used in simulated shootings. Such information could potentially help investigators better determine the path a bullet took through a victim’s body, Sascha Dammeier, a Tuebingen researcher and first author on the study, told GenomeWeb.
The researchers have not yet been able to generate solid organ-specific profiles for bullets collected at actual crime scenes. However, Dammeier noted, they have in some cases been able to aid in incident reconstructions by identifying organ-specific proteins present on bullets and other material.
He added that they have patented the approach and are reaching out to forensic labs to see if they might find it useful, although, he said, he doesn’t expect it to become a steady business.
The project came about when the university’s proteomics center moved into a building that was formerly occupied by forensic medicine researchers, Dammeier said.
"Just by chance there were two people left [in the building] who lecture in forensic medicine, and they came over one day and said to me, 'Proteomics, what is that? What can you do?'" he recalled.
"So, I explained it to them and they were very interested," he said. In particular, they were interested in proteomics’ potential to help with reconstructions of shootings involving more than one bullet.
In such cases, the forensics researchers told Dammeier and his colleagues, it can be difficult to determine which path which bullet took through a body.
"Often, when there are two people involved, they both will say, 'Oh, the other one did it. It was not my bullet that was the lethal one,'" Dammeier said.
While DNA is commonly used for various types of forensic analysis, it wasn't a suitable molecule for answering such a question where organ-specific profiles are required. Better, said Dammeier, would be mRNA, but mRNA degrades quickly, especially under the sort of conditions a fired bullet might be exposed to, and so was also less than ideal.
Forensic researchers had also tried looking at protein markers, but they were typically doing so using immunofluorescence assays that, Dammeier said, were not particularly well suited to the task. "They can only look at one marker at each time and it is quite [challenging] to do this on projectiles," he said.
As a result of these limitations, he and his colleagues turned to mass spec to generate organ-specific protein expression profiles that could aid investigators in determining bullets' paths.
They begin by building their profiles with bovine tissue, taking 79 samples from five organs — heart, kidney, liver, muscle, and lung — and manually pressing metal through them to simulate a hit from a bullet. They then digested the protein captured by the metal directly on the surface and measured the resulting peptides in three-hour LC-MS/MS runs.
Using this data, they developed a variety of classifiers for identifying which organ the captured proteins came from, finding that the multinomial naïve Bayes classifier using the top three most discriminating proteins per organ gave very good performance. Simply looking at whether these top three proteins are present or not allowed, the researchers identified the involved organs with an accuracy of 99.1 percent — and when protein expression data was included, this number rose to 99.8 percent. Ultimately, they determined that use of the top five proteins per organ provided the most effective classifier.
While the researchers had success in applying the approach in this experimental setting, using it in situations more closely approximating an actual shooting proved more challenging. For instance, they next did a series of shooting experiments in which they fired bullets into four bovine organs — liver, lung, kidney, and heart — embedded in gelatin.
Using the same LC-MS/MS method, they identified proteins captured on these bullets and, taking the top five per organ, applied their classifier to determine which organ they had passed through.
In this case, they were able to accurately identify the organ in the case of two of the four shooting samples — those that passed through liver and lung. The heart sample they misidentified as muscle (though, as the authors noted, this was not technically a misidentification, given that the heart is a muscle), while the kidney sample they misclassified as lung.
This partial success in hand, they then turned their method to analysis of a recent local murder case in which investigators had not been able to fully determine the paths of the multiple bullets through the victim.
Using their method they looked at a total of five bullets, two (exhibits 1-12 and 1-13) found inside the car where the shooting took place and three found in the street (exhibits (1-14, 1-17, and 1-3). Via mass spec, they were able to identify on the bullets 1-12, 1-13, 1-14, 1-17, and 1-3, respectively, 269, 180, 33, 34, and 84 proteins.
The Tuebingen researchers were not, however, able to discriminate between the different organs hit using these protein identifications. As they noted, the bullets had likely penetrated multiple organs, making contamination an issue. Additionally, the bullets likely hit organs not covered by the group’s bovine experiments.
Changing course, the researchers moved from using their classifiers to trying to match the proteins they were able to identify to the top organ discriminating profiles. In this they were more successful, finding, for instance, that bullet 1-12 was the only projectile to contain several of the top discriminating proteins for liver, which helped investigators match it to one of the paths of penetration through the body. Additionally, they showed that bullet 1-13 had several heart- and lung-specific proteins but no liver-specific proteins, which suggested its most likely path, as well. They were also able to determine the likely path of bullet 1-3.
"So, we found that the classification works for the experimental setting, the shooting setting, but maybe it does not work for the real case scenario, because you have lots of contamination not only by cloth but by skin, muscle, and so on," Dammeier said.
He noted that he and his colleagues have since applied a similar approach to several other cases, first trying their classifier and then moving to the second approach in which they used the presence of specific organ-associated proteins to help reconstruct shootings.
The Tuebingen researchers were particularly well positioned to tackle the problem, Dammeier said, given a previous industry collaboration they had taken part in looking at the proteomic profiles of explants taken out of patients. For that work "we had to look at the tissue's reaction to the [explant's] surface," he said. "We were looking at surface proteomics, so that's why we were a little familiar already with this sort of analysis."
Next, Dammeier and his colleagues plan to do a follow-up study applying to technique to knives. "That might be a little more useful, at least here in Europe where we don't have a lot of shootings," he said.