Forensic Proteomics: Going Beyond DNA Profiling

Detecting a single protein is useful, but because there are so many different types of proteins—a few thousand in a single bacterial cell and tens of thousands in human tissue—it would be expensive and time-consuming to make an antibody to detect every protein one by one.

Proteomics is different because it looks at all the proteins in a sample at the same time. To do this, researchers relax many tightly coiled proteins using chemicals called denaturants. Next, they use enzymes to cut the relaxed proteins into short amino acid chains known as peptides. Mass spectrometers analyze the peptides, tracking the number and type of peptide signals they detect.

The researchers assemble all the peptide signals to figure out how many point to a protein, organism, or tissue of interest, instead of detecting signals from just a single protein and ignoring everything else.

Gathering so much data makes proteomics ideal for forensics, where a sample doesn’t come with much background information to hint at which antibodies to use or whether to use antibodies at all.

“Suppose you think there’s something there, but you don’t know what it is,” said Merkley. “In that case, testing for lots of things at once is the ideal scenario.”

This shotgun testing approach sounds difficult; however, acquiring all those signals is the easier part. Analyzing the signal data to know what it means—and so it can be reliable in a courtroom—is at the heart of the book.

Making a Case for Forensic Proteomics
To be admitted in federal court, scientific evidence must meet the Daubert standard, a rule for scientific testimony that assures experts validate forensic methods. Publishing is one of the best ways to show that a technique is defensible in court because it subjects the method to peer review, which is one measure of acceptance in the scientific community.

Legal admissibility also includes showing how a method could be falsified and refuted. DNA is powerful because it relies on well-known probabilities and statistics to identify individuals with as much as one in 575 trillion specificity. Developing these precise probabilities for forensic proteomics is key to achieving the same type of legal prowess.

To illustrate the need for probabilities, Merkley gives the example of a protein toxin called ricin that shares a peptide with a protein from a Central American aquatic snail. If an investigator finds only that peptide, they can’t necessarily know they have identified ricin, but they can be more certain if they find 10 other peptides from ricin that aren’t found in the snail.

“Multiply that problem hundreds of thousands of times across all the organisms on Earth,” says Merkley, “and it becomes clear each observation doesn’t have the same weight for saying a certain thing is there.”

Further complicating an already tricky problem: There isn’t a database that identifies every peptide sequence in the millions of proteins in every organism that has ever lived. Researchers must therefore quantify how frequently—or rarely—peptide combinations occur, consistent statistical standards that only experts can create.

Merkley teamed with Kristin Jarman, a PNNL statistician, for the last chapter of the book to suggest statistical methods that he hopes to see the scientific community expand.

“I would like to see more efforts to codify standards,” says Merkley. “My hope from the book is that it develops a strong community to find the best way to develop this tool together.”