Eleanor Watts

DPhil, MPH, BSc (Hons)

Eleanor L. Watts

Postdoctoral Fellow

PLCO (Learn more about this study)

2023-0061

Jun 12, 2023

Quantifying the proteome: exploratory associations with hepatocellular carcinoma

Liver cancer is a common malignancy and the third leading cause of cancer death worldwide1. The most common histological type of liver cancer is hepatocellular carcinoma (HCC). Risk factors for HCC include chronic hepatitis B, hepatitis C, excessive alcohol use, and diabetes/obesity/non-alcoholic fatty liver disease2. However, our understanding of the mechanistic pathways through which these risk factors are associated with HCC remains incomplete.

In the last decade, rapid developments in proteomics technologies have enabled the measurement of thousands of proteins using a single analytic platform. These advances have enhanced our ability to study the mechanisms and pathways potentially underlying etiological relationships and may enable identification of candidate therapeutic targets and markers of disease detection and prognosis3-6. One leading proteomics platform, by the company Olink, is currently able to detect 3,000 proteins, with high technical reproducibility and specificity3, 7, 8. This platform has been previously used to identify intriguing associations with cardiovascular disease, diabetes, and lung cancer3, 4, 9.

To take advantage of these new capabilities, we propose a pilot HCC study in PLCO that will evaluate associations between proteins and risk of HCC and provide data on platform reproducibility and intraindividual variability in protein levels over time. If results are promising, subsequent proteomics efforts could be expanded through CGR. With repurposing of existing equipment and some purchases, the per sample costs can be reduced to less than half of the price. These subsequent efforts could include large cancer studies (e.g. expansion to other cancers, increased sample size for the HCC proposal) and detailed methodological investigations (e.g. evaluation of sample handling effects). Quantifying the proteome in large studies would be a powerful tool to increase our understanding of disease, help to advance DCEG’s cancer research mission and benefit transdisciplinary research across branches. These data will also complement PLCO’s large collection of phenotypic, genomic and metabolomic data.

In brief, the study will include a total of 264 samples over 3 proteomics plates (88 samples per plate). This will include a nested case control study of 120 HCC cases and matched to 120 controls. Selection for cases will be based on those with serum available with the smallest number of prior freeze thaw cycles. Control selection be based on those included in previous metabolomics, hepatitis and sex hormone and liver cancer nested case control studies. To investigate measurement repeatability, we will include blinded quality control samples using triplicate samples from this control group (7 triplicate control sets). We will also include an additional 10 blood samples from year 1 from the controls to investigate the median intraindividual variability of proteins over time.

References
1. https://gco.iarc.fr/today
2. doi:10.3748/wjg.14.4300
3. doi:10.1038/s41467-021-22767-z
4. doi:10.1101/2022.06.17.496443
5. doi:10.1016/j.cell.2020.10.037
6. doi:10.1038/s41586-018-0175-2
7. doi:10.1126/sciadv.abm5164
8. doi:10.1101/2022.02.18.481034
9. doi:10.1101/2022.07.31.22277301

Aim 1: Using blinded quality control samples, we will calculate the intraclass correlation coefficients (ICCs) and the coefficient of variation (CVs) for each measured protein from the 7 triplicate serum samples from our study controls (drawn from 7 controls).
Aim 2: Calculate the median ICC from repeat blood samples 1-year after study baseline in a subset of 10 controls to investigate temporal variability of proteins (10 samples)
Aim 3: Estimate associations of proteins with HCC risk using conditional logistic regression (120 HCC cases matched to 120 controls).

We estimate that we will have 80% power to detect an OR per 1 SD increase of 1.97, using a conservative Bonferroni correction for multiple comparisons. We have previously observed similarly large magnitudes of associations for other proteins with liver cancer (e.g., IGF-I: OR per 1 SD=0.32, 95% CI 0.26-0.39)1, 2. The minimum OR required to detect an association for a p-value <0.05 is 1.44.

If the Olink technology is demonstrated to have acceptable reproducibility and temporal variability we hope that this foundational work might lead to purchasing Olink equipment, which will more than halve costs and enable the expansion to other PLCO samples in the future.

References
1. doi:10.1002/ijc.33555
2. doi:10.1158/0008-5472.Can-20-1281