In a recent article published in the journal BMJ Oncology, researchers developed a novel proteome/liquid biopsy-based diagnostic test to detect less abundant plasma proteins, i.e., biomarkers for various cancers in human organs.
The researchers were pursuing the development of multi-cancer tests that might overcome the lower sensitivity of current protein assays and nucleic acid tests in detecting protein biomarkers of early-stage cancers.
Early cancer detection with high accuracy is vital, given the increasing global incidence of cancer. It can enable prompt initiation of curative treatments.
Nearly 60% of cancer-related deaths occur in the absence of reliable tests for the detection of cancers in the early stages. In addition, currently used tests, such as mammography, colonoscopy, etc., are invasive, costly, and less accurate for early-stage cancers.
Genomics-based liquid biopsy can enable the development of a blood test for simultaneous screening of multiple cancers. However, currently, these tests are expensive (priced over US$500) and have shown less than 50% sensitivity for early-stage cancers.
To date, the potential use of plasma proteins in proteome as biomarkers for cancer has remained challenging due to the complexity of the proteome and lower sensitivity in detecting low-abundance proteins.
Nonetheless, many such blood-based protein biomarkers have shown the ability to detect and monitor cancers early. Examples include cancer antigen 15-3 (CA 15-3), CA 27.29 for breast cancer, carcinoembryonic antigen (CEA) and CA 19-9 for colorectal cancer, and alpha-fetoprotein for liver cancer.
In the present study, researchers developed a novel plasma proteome-based test and showed its potential to detect early-stage solid tumors.
They collected plasma samples from 440 cancer patients and healthy individuals, with the former diagnosed with one of 18 different solid tumors (excluding melanoma), representing nearly all human organ cancers.
They used Olink’s proximity extension assay (PEA) technology to measure proteins in these plasma samples. Technically, this assay employs antibody-based detection to assess the levels of 3072 target proteins in plasma.
The team performed two statistical steps separately for male and female samples to calculate a probability-based score for each step. While the first step involved a search for a limited number of proteins that could identify any cancer in its early stages, the second step classified each type of cancer to find a cancer-specific tissue of origin.
The team selected the proteins exhibiting the highest number of non-zero coefficients using the L1 penalty to 100 bootstrap samples of the original dataset. It prevented the simultaneous selection of correlated biomarkers.
They assessed the performance of all biomarker panels, expressing results as the area under the curve (AUC) of the receiver operating characteristic curve. Furthermore, they used the leave-one-out method to evaluate the performance of the biomarker panel with selected features.
Most study participants were asymptomatic and were diagnosed with early-stage tumors after they underwent routine medical check-ups; in addition, they all were treatment-naïve. Of 3,071 proteins analyzed from their samples, 2,785 passed the quality measurements.
Many protein pairs displayed a high positive correlation. From these, the researchers selected the most informative proteins for cancer diagnosis. While these plasma proteins could differentiate cancer samples and distinguish between different types of cancers with high accuracy, the protein-cancer association varied significantly between males and females.
For the male and female cohorts, 80% and 83% of proteins with a p-value below 0.05 had no significant difference in females and males, respectively. For a more stringent p-value threshold of 0.001, these values increased to 97.8% and 99.1% for males and females, respectively.
When trying to identify sex-specific protein sets to identify cancer, the researchers expected to see an increasing performance of the model by AUC with more proteins to capture different cancer populations. Intriguingly, AUC did not improve further in any model after it reached the threshold of 10 proteins.
Each protein in the panel alone had a low to medium detection accuracy; however, in combination with other proteins as a panel, they achieved a very high accuracy in the detection of early-stage cancers.
The overall sensitivity of this test was 90% and 85% among males and females, respectively, at 99% specificity.
The present study evidenced that proteins present in low concentrations in the plasma proteome were the most valuable as biomarkers for early-stage cancer detection, and cancer protein signatures were sex-specific. This knowledge could present new avenues of research in proteomics and cancer biology.
This knowledge also paves the way for a single, inexpensive, and accurate screening test for multiple cancers, facilitating cancer screening in the general population.
However, there is a need for greater validation of the test in larger cohorts to establish its reliability and generalizability.
Nonetheless, the implementation of such a test in healthcare systems could reduce cancer-related health and economic burdens. If the test became integrated into routine check-ups, it could reshape cancer screening guidelines in the future.