Discussions
In this study, we showed that a measurement of limited set of plasma proteins could classify cancer samples from normal and differentiate different cancers. This finding is the foundation for a multi-cancer screening test for the early detection of 18 solid tumours that cover all major human organs of origin for such cancers at the earliest stage of their development with high accuracy. It is important to diagnose cancer at very early stages where curative treatments are achievable with surgery and available treatments. Additionally, for the first time to our knowledge, we found compelling evidence that the cancer protein signatures are most likely sex-specific for all cancers.
Our study also showed that biological signals for early-stage cancers are much more evident in the low-concentration part of the human plasma proteome. It was also promising to observe that a set of proteins could differentiate all cancers from normal and sensitive to detect unseen cancers.
In our study, we analysed a range of proteins found in classical cancer pathways. However, we discovered that only a very small number of these proteins could be used as biomarkers for early-stage cancer. In contrast, many proteins that were effective biomarkers for early-stage cancer were found at low concentrations across the entire plasma proteome. This finding may be due to the fact that most of our knowledge about the role of proteins in cancer pathways comes from studies of transcriptome at the tissue level in advanced stages of cancer, and the expression of proteins at the mRNA and protein levels do not always correspond. In addition, the concentration of proteins in tissue and plasma may not be strongly correlated. Finally, our samples were mainly from early-stage cancers, where classic cancer pathways may not be highly active. This finding has major implication for developing the next generation of diagnostics highlighting the role low-abundance protein in early detection of disease.
The proteome-based diagnostic test showed promising performance compared with other technologies such as circulating tumour DNA tests23 by significantly outperforming existing multi-cancer screening tests in detecting cancer across all stages (I, II, III) and among all types of cancers. At the specificity of 99% and in stage I of cancer, our test had a sensitivity that was much greater than Galleri24 and CancerSEEK25 tests. Additionally, our study demonstrated ability of our ‘best-test to achieve much higher accuracy in identifying the tissue of origin of cancers in each sample in comparison to other tests. At the cancer-specific level, all our best-tests were more accurate than other available screening tests. Among the four screening tests that have received the highest recommendation (level A) from the US Preventive Services Task Force (colonoscopy for colon cancer, pap test for cervical cancer, mammography for breast cancer and low-dose CT scan for lung cancer), only colonoscopy and low-dose CT scan had an accuracy of above 90% for cancer detection. However, the sensitivity of our test for detecting early-stage cancer was still higher than the sensitivity of these tests.
Over the past decade, mRNA large-scale sequencing has provided a comprehensive view of gene expression in specific tissues, revealing the proteins that are present in different organs of the human body. The Human Protein Atlas is a useful resource for understanding mRNA and protein expression in multiple healthy tissues.26 However, it is important to note that tissues are typically composed of complex assemblies of distinct cells that may have different functions and developmental histories. Increasing amounts of information about RNA and protein expression in specific cell types is now becoming available for the individual cells that make up tissues and organs. A challenge in using protein detection for liquid biopsy is that cancer-specific protein biomarkers may be present at ultra-low levels in the blood.9 This is because proteins that are present at high concentrations in the blood of healthy individuals are unlikely to be significantly increased in patients with early stages of the disease or at early recurrence. The long history of plasma proteome analysis by mass spectrometry show that even proteome coverage was increased from several hundred proteins 30 years ago to more than 5000 proteins based on latest development in chromatography separation technique and data independent acquisition type of acquisition.27 Still the major problem of cheap and reproducible sample preparation protocols and reliably measuring proteins after first thousands of most abundant proteins prevent development of early stage multi-cancer test by mass spectrometry at acceptable price per sample and general population scale. Thus, assays with greater sensitivity for biomarker proteins that are normally present at very low or undetectable levels in the blood may enable the detection of cancer at an earlier stage of the disease or even at premalignant stages. Our test is based on sensitive proximity assays that require the simultaneous binding of three separate antibodies. This ability to analyse plasma proteome profiles deeply and consistently allowed us to focus our attention on very low-abundant proteins, which we found to be the most precise and accurate biomarkers of early stages for all the cancers studied in our study. Advancing the PEA technology to measure ultra-low protein concentrations will provide better opportunities to detect and classify cancers at a very early stage and even at the precancerous stage.28
Our new generation protein-based plasma test has shown high sensitivity in detecting a variety of early-stage tumours in asymptomatic patients, making it a strong candidate for use as a population-wide screening tool that is not currently achievable with existing tests or techniques. Its high specificity can help alleviate concerns about causing harm to patients, and its low cost allows for widespread implementation. To be suitable for large-scale use, a screening test must have high sensitivity and the ability to reduce mortality and morbidity, as well as acceptable for healthcare system cost. In the case of cancer screening, it is also essential for the test to have high specificity to avoid causing undue harm to patients. Our test exhibits these desirable qualities, making it a promising option for cancer screening. We expect that the combination of lower cost and higher accuracy in our test will facilitate its integration into the healthcare system and eventual inclusion in routine annual check-ups. Early detection of cancer has the potential to greatly reduce the societal burden of both health and financial costs. In fact, implementing such interventions can not only be cost-effective but can also result in cost savings for society.
In our study, notable sex-specific differences in cancer detection emerged, necessitating a deeper exploration. The types of cancers in our pan-cancer screening inherently differ between males and females due to the presence of sex-specific cancers like prostate or ovarian cancer. Beyond this, certain proteins are exclusive or more predominant in one sex, affecting detection accuracy. Additionally, the overall distribution and abundance of proteins vary between males and females, with the relationships between these proteins also being sex-specific. Recognising these inherent biological differences highlights the potential benefits of employing gender-tailored biomarker panels, which might enhance detection accuracy. This approach underscores the significance of personalised medicine in contemporary oncology, ensuring diagnostics are attuned to the unique biological signatures of each gender.
Our approach has major strengths, including the total number of proteins measured and accuracy of such measurements across all measured proteins down to very low abundant proteins, the focus on early-stage tumours, the number of studied cancers that represents all major organs of unmet needs included in the study.
Limitations should also be considered in the interpretation of our study findings. The size of the cohort used in this study was relatively small. While we aimed to capture a diverse range of cancers, the limited sample size may restrict the generalisability of our results to larger populations. Therefore, it is important to validate our test in larger population cohorts to ensure its robustness and reliability across different demographic groups and geographical regions. Another limitation of our study was the lack of comprehensive information on key patients’ comorbidities (eg, diabetes, hypertension, obesity). Due to the retrospective nature of the study and the limitations in data availability, these variables were not captured for all cancer subtypes and normal individuals in our dataset, and we were not able to assess their effect on the performance of our panel. Comorbid conditions may introduce additional variability and complexity to the analysis. The influence of these comorbidities on the performance of our test and its accuracy in detecting early-stage cancers needs further investigation. Validation in a cohort with a more diverse range of comorbidities will provide a more comprehensive understanding of the test’s performance in real-world clinical settings. Another limitation of our study is the uneven stage distribution of cancer cases, as highlighted in table 1. Our emphasis on sourcing treatment-naive samples, primarily from patients diagnosed during routine check-ups, inadvertently led to an over-representation of specific disease stages. Notably, some cancers had only stage II or III representation, potentially missing insights from other stages. This approach, while offering a unique subset of samples from the biobank, might not comprehensively represent the broader population of cancer patients across all stages. Furthermore, while our study focused on a comprehensive set of proteins and achieved accurate measurements across a wide range of proteins, there may still be limitations in terms of proteome coverage. Despite advancements in proteomic techniques, the complete coverage of the plasma proteome remains a challenge. Certain low-abundance proteins may not have been captured in our analysis. Improvements in proteomic technologies and sample preparation protocols are required to enhance the sensitivity and coverage of the proteome, allowing for a more comprehensive assessment of biomarkers for early-stage cancer detection. Additionally, our study mainly focused on early-stage cancers, where classic cancer pathways may not be highly active. This narrow focus may limit the generalisability of our findings to advanced stages or metastatic cancers. Future studies should aim to explore the performance of our test across a broader spectrum of cancer stages and subtypes to evaluate its effectiveness in different clinical scenarios. Finally, while our test demonstrated promising performance compared with existing technologies and screening tests, it is essential to emphasise the need for independent validation in an external cohort. Validation in an independent population will provide further evidence of the test’s accuracy, sensitivity and specificity. It will also help assess its performance in diverse patient populations, accounting for variations in genetic backgrounds, environmental factors and healthcare settings. Robust validation studies are necessary to establish the clinical utility and reliability of our test before its widespread implementation in routine cancer screening programmes.