With great interest, we read the article by Law and colleagues (1). They describe the proteomic landscape of liver metastases of a patient cohort suffering from pancreatic cancer. Unsupervised classification showed four molecular subtypes in liver metastases. These subtypes correlate to chemosensitivity and prognosis. One limitation of subtyping in pancreatic cancer is the lack of consensus of previous transcriptomic subtypes, unknown functional relevance, and the need for correlation with IHC and/or phenotypic histology (2). This has resulted in lack of implementation in clinical practice. The proteomic subtypes described in this study can hopefully assists in validation of subtype-specific proteins. However, some key points of the study should be further discussed.

Protein identification by mass spectrometry was performed and data were filtered for proteins identified in the whole cohort. This boosts qualitative data for comparisons, however reduces variability and thus might exclude important low abundant processes. Moreover, samples were gathered upon death, which induces a complicated variable: progression of disease and subsequently heterogeneity of treatment, resistance, and patient characteristics. Pancreatic cancer shows resilience against treatment regimens, a plasticity that is not taken into account within this cohort. Therapeutic pressure can induce clonal selection, therefore reduces clinical relevance of identified