Background
A team from the University of Liverpool have reported their findings from a remarkable large-scale proteomics study into early detection and risk prediction for lung cancer. Using Olink® Explore 3072, they measured 2941 proteins in 496 Liverpool Lung Project plasma samples, including 131 cases taken 1–10 years prior to diagnosis, 237 controls, and 90 from subjects sampled longitudinally at multiple times. Differentially expressed proteins were subsequently modelled for lung cancer prediction and the findings were then validated and extended using data generated by the UK Biobank Pharma Proteomics Project (UKB-PPP)
Outcome
For samples taken 1–3 years pre-diagnosis, 240 proteins were significantly different in cases v controls, while for those taken 1–5 years pre-diagnosis, 117 of these and 150 further proteins were identified, mapping to significantly different pathways. Four machine learning algorithms gave median AUCs of 0.76–0.90 and 0.73–0.83 for the 1–3 year and 1–5 year proteins respectively. Using the proteins that overlapped between the two studies, the findings were then externally validated using the online dataset from UK Biobank. This gave AUCs of 0.75 (1–3 year protein model) and 0.69 (1–5 year protein model), with AUC 0.7 up to a remarkable 12 years prior to diagnosis. These models were independent of major lung cancer confounders such as age, smoking duration, cancer histology and the presence of COPD.
In addition to the high-performance of the protein signatures identified, the study also revealed some valuable biological insights into the proteomics of lung cancer. Most crucially, while there was partial overlap between biomarker proteins selected using longer and shorter pre-diagnostic time points, well over a hundred proteins were unique to each sample set. This suggests that different biomarkers for occult early tumor detection and inherent cancer risk can both be identified using this plasma proteomics approach. This conclusion was strengthened by clear differences in the molecular pathways associated with the two types of markers. This is an important distinction, because while one type of biomarker facilitates earlier diagnosis, the other contributes to the stratification of high-risk individuals for screening, or targeted preventative measures.
