GRAIL develops technology for the muli-cancer early detection (MCED), and has developed an MCED test, Galleri, that detects cancer-derived cell-free DNA (cfDNA) fragments in plasma. Studies of the UK Biobank proteomics data (Papier et al, 2024, Nature Communications) and in other cohorts have suggested that plasma proteins may have the potential to additionally serve as biomarkers for early cancer detection. However, it is not yet clear whether proteins could robustly detect cancer in a clinical implementation, or whether they could complement cfDNA for cancer detection.
To develop plasma proteins as a cancer biomarker, it’s critical to identify proteins that exhibit consistent performance as a cancer biomarker across a broad range of patient populations (age cohorts, sex, and ethnicity). We propose to use the UK Biobank cohort to determine whether proteins associated with cancer status display robust detection of cancers across different age cohorts, ethnicities, and sex and whether a model trained on these cancer biomarker proteins produces reliable results on the UK Biobank cohort across these potential confounders.
If we are able to generate a model predictive of cancer status that is robust to these confounders, we will assess whether the cancer predictions made by this model complement cfDNA-based cancer predictions made by GRAIL’s Galleri test, either by identifying new cancers not identified by cfDNA, or by identifying “false positive” samples that display cfDNA signal that is not derived from a cancer. If complementarity is identified, there may be an opportunity to incorporate protein analyses into future iterations of GRAIL’s cancer detection and monitoring products, improving cancer detection and treatment.