Journal: Molecular Oncology

Article Title: Profiling of extracellular vesicles of metastatic urothelial cancer patients to discover protein signatures related to treatment outcome

doi: 10.1002/1878-0261.13288

Figure Lengend Snippet: Protein profiling of extracellular vesicles in relation to progression‐free survival. Extracellular vesicles (EVs) isolated from plasma samples from metastatic urothelial cancer (mUC) patients at day 21 were subjected to proximity extension assay (PEA) protein profiling with the Oncology II ® assay (see Section ). (A) Rank regression analyses of protein signatures in EVs related to progression‐free survival (PFS) [long (green) vs short (red)] of the study population. The qlucore software rank regression tool was used with P ≤ 0.05, and protein signatures were sorted without ( left panel ) or with ( right panel ) the number of EVs applied in the PEA profiling as elimination factor (see Section ). For description of the processing of the PEA data prior to the Qlucore analyses with respect to samples showing protein expression below lower limit of detection (LOD) see Sections and . The PEA data were also analysed with the XGBoost integrated tool of the qlucore software (see Section ). The proteins sorted out by this approach and that also were revealed by univariate rank regression analyses are indicated with (*). (B) The normalized protein expression (NPX) values of individual proteins from (A) were analysed in EVs from plasma of patients with short (≤ 138 days) or long (> 138 days) PFS ( top panel ) or after normalising for the number of EVs used in the PEA profiling ( bottom panel ). Only proteins that were statistically significant using t ‐test (see Section ) at P ≤ 0.05 and ≤ 0.06 are shown. Bars represent standard deviation (SD) values. Proteins that only showed significant association with PFS when non‐normalised PEA data were used are presented in Fig. . Please note that pat. #114 was excluded in these analyses. For LOD of the proteins, see Table . (C) The expression of individual proteins from the PEA analyses was plotted with linearised values against PFS in linear regression analyses (see Section ). The table indicates the Pearson correlation coefficient alongside P ‐value for non‐normalised samples. Please, note that pat. #114 was excluded in these analyses. Presented data in A–C are from one PEA profiling of one biological isolate of EVs from plasma of the individual patients. (D) Western blot profiling of SYND‐1 and GZMH in EVs isolated from plasma samples of mUC patients at day 21. SYND‐1 and GZMH were profiled without normalising for the amount of EVs in the different samples. The fold expression of SYND‐1 relative to pat. #107 is given with (fold norm) or without (fold) normalisation for the number of EVs analysed. The densitometric quantification of SYND‐1 (without normalisation) is presented in Fig. . Data shown are from one western blot analysis of one biological isolate of EVs from the individual plasma samples of the patients.

Article Snippet: The PEA data were also explored using a machine learning algorithm XGBoost (Extreme Gradient Boosting, Qlucore integrated tool; see https://xgboost.readthedocs.io/en/latest/ #) to build classifier models.

Techniques: Isolation, Ii Assay, Software, Expressing, Standard Deviation, Western Blot

Journal: Molecular Oncology

Article Title: Profiling of extracellular vesicles of metastatic urothelial cancer patients to discover protein signatures related to treatment outcome

doi: 10.1002/1878-0261.13288

Figure Lengend Snippet: Profiling of extracellular vesicles to reveal putative protein signatures in relation to treatment response. Extracellular vesicles (EVs) isolated from plasma samples from metastatic urothelial cancer (mUC) patients at day 8 post‐treatment were subjected to proximity extension assay (PEA) protein profiling with the Oncology II ® assay as described in Section . Please note that for pat. #110, fractions 7–10 were analysed, while for all the other patients, fractions 6–10 were examined. (A) Rank regression analyses of protein signatures ( P ≤ 0.05) in EVs related to best response of the patients evaluated by computerised tomography (CT) (Fig. B) are presented. The analyses were carried out as in Fig. with the number of EVs analysed applied as an elimination factor (see Section ). The PEA data were also analysed with the XGBoost integrated tool of the qlucore software (see Section ) at day 21. Star (*) indicates that FASLG, which was revealed by univariate analyses, also was identified with this method. (B) The linear expression of indicated proteins in individual EV samples at day 8 and 21 was plotted against best CT response without normalisation for the number of EVs analysed. The line indicates results from the linear regression analyses. The Pearson correlation coefficient is given alongside the P ‐value. The lower limit of detection (LOD) and RIPA negative control values are presented in Table . Please note that pat. #114 was excluded in these analyses (see Section ).

Article Snippet: The PEA data were also explored using a machine learning algorithm XGBoost (Extreme Gradient Boosting, Qlucore integrated tool; see https://xgboost.readthedocs.io/en/latest/ #) to build classifier models.

Techniques: Isolation, Ii Assay, Tomography, Software, Expressing, Negative Control