Mass spectrometry is widely used by chemists to identify unknown molecules in a (typically liquid) sample. Identification is provided by the mass to charge ratio of fragments of these molecules.
It provides much useful identifying information unavailable to other types of spectrometry. However, the huge variety of molecules present in blood has hindered the use of mass spectrometry for clinical diagnoses.
Mikkel West-Nørager (Statens Serum Institut, Denmark) and coworkers have evaluated the feasibility of tackling this challenge. They have taken steps to show that, with further development, mass spectrometry may be capable of providing a clinical cancer diagnosis.
Why choose mass spectrometry?
There are many clinically-proven protocols for cancer diagnosis. Why would anyone want to use mass spectrometry?
The big reason is that mass spectrometry can in theory conclusively prove the presence, identify, and report the concentration of all of the small peptides and proteins in an unknown sample, such as blood. No other diagnostic protocol can do this.
Why is this important? Cancer is often diagnosed by an abnormal amount of a particular peptide or protein in blood.
Additionally, traditional cancer diagnoses are specific to one type of cancer. In theory, mass spectrometry can instead be used as a universal cancer screening.
The big problem with using mass spectrometry for cancer diagnosis is, paradoxically, its advantage: the huge amount of data it provides. The identifying information provided by mass spectrometry on a peptide or protein is mixed in with identifying information on all of the other molecules present in the sample.
This mixup makes finding the "signature" of one particular molecule extremely difficult to extract when there are many other molecules (of widely varying concentrations) present in the sample. Envision hunting for a 1992 penny in a large barrel of coins.
The blood samples.
Towards addressing this challenge, the scientists obtained blood samples from women admitted to a Copenhagen hospital, all about to receive surgery for suspected ovarian cancer. All blood samples were taken less than 15 days prior to surgery, and samples were used within 6 hours.
Patients with borderline ovarian cancer and benign tumors were also included in the study as controls. For analysis, blood samples were analyzed randomly with no knowledge of their diagnoses.
Searching for useable data.
As discussed previously, mass spectrometry of a complex mixture provides tons of information, but it is difficult to extract the information you want from everything else in there. In this case, the scientists found approximately 33,000 data points in each blood sample.
A large reduction in data points was achieved by eliminating data that was not far outside the noise level. This produced 117 data points for which a mass to charge ratio could be determined.
Testing the mass spectrometry cancer diagnosis method.
The scientists performed extensive statistical analyses on their blood sample characterization method. An important result here is that the technique was able to discriminate cancer from non-cancer blood samples 68% of the time, with 21% false positives and 44% false negatives.
How does this compare with established methods of detecting ovarian cancer? The established method gives only 3% false positives and 32% false negatives.
It is clear that mass spectrometry is not yet ready to compete with traditional clinical cancer diagnoses. However, the fact that any data patterns were found with mass spectrometry that correlate with cancer is promising (in the long term) for rendering mass spectrometry into a universal clinical cancer diagnosis tool, not limited to one specific type of cancer.
for more information:
West-Nørager, M.; Bro, R.; Marini, F.; Høgdall, E. V.;
Høgdall, C. K.; Nedergaard, L.; Heegaard, N. H. H.
Feasibility of serodiagnosis of ovarian cancer by mass spectrometry.
Anal. Chem. 2009, 81, 1907-1913.