31 Context, background, aims, and outline of the thesis The use of regression models to establish the prognostic value of a biomarker It is generally accepted in clinical prediction modeling that variable selection should be based on clinical insight and knowledge extracted from the literature (25). Yet, to avoid overfitting (26), it is advised to use the ‘one in ten rule’ when using traditional clinical prediction modelling strategies, such as logistic regression and survival models (27). This rule states that for every 10 events in a dataset one variable can be considered in a model (25, 28, 29). Besides clinical insight, literature knowledge, and experience, candidate variable selection for the final model should be based on the data and data analysis (27). It is advised to combine (30) or exclude highly correlated (31-33) variables, restrict inclusion based on distribution (30, 33), and exclude in case of a large number of missing values (30, 33). Once the candidate variables have been identified, a definitive selection of variables needs to established for the final model (27). This can be done using the full model approach (inclusion of all candidate variables) (33), or formal variable selection methods (e.g. forward and backward elimination) when there is confusion or uncertainty regarding which variables to consider in the final model (27). MET ectodomain shedding In addition to phosphorylation, the activity of MET signaling can be regulated by controlling the number of receptor molecules present on the membrane (34). A process called presenilin-regulated intramembrane proteolysis (PS-RIP) degrades MET through sequential proteolytic cleavage. First, the receptor is cleaved within its juxtamembrane domain by membrane metalloproteases. This process is independent of ligand stimulation and requires no kinase activity. Consequently, a soluble MET N-terminal fragment (MET-NTF) is shed into the extracellular space, an occurrence referred to as ectodomain shedding. ADAM metalloproteases 10 and 17 (ADAM10 and ADAM17) drive MET-NTF shedding, as it is inhibited upon their genetic ablation (35, 36). Immediately after shedding, the remaining membrane bound 55 kDa C-terminal fragment (MET-CTF) undergoes a second cleavage, which is performed by the γ-secretase complex. The resulting 50 kDa intracellular domain of MET (MET-ICD) is degraded by the proteasome upon its release into the cytosol. The MET-CTF fragments that escape γ-secretase cleavage are subjective to lysosomal degradation (Figure 1) (37, 38). 2
RkJQdWJsaXNoZXIy MTk4NDMw