Despite the clinical importance of discrimination, reliable diagnostic biomarkers are still lacking.
Despite this clinical importance, reliable biomarkers for differentiation between SQCLC and HNSCC metastasis in the lung are currently lacking; this may be due, at least partly, to the overlapping aetiology, morphology and biology of these tumour entities.
We reasoned that this proteomic approach, rather than genomic and transcriptomic studies, may identify a suitable biomarker panel and may also provide some more general insights into the biology of these tumours.
However, diagnostic biomarkers are so far lacking, owing to the large number of morphological and genomic features that are shared by these tumour entities. Because the two tumour types have not been systematically compared at the proteome level, we wished to explore the possibility of identifying proteomic diagnostic biomarkers.
All patients were treated by surgery and 40 of them had received additional adjuvant chemo radiotherapy. None of the patients had received neoadjuvant therapy or primary chemoradiotherapy.
The MS analysis of each sample was performed in duplicate, and data analysis was performed with the software packages MaxQuant, Perseus and R. Whiskers indicate minimum and maximum. Shown are the first two principal components, accounting for 8. Notably, the correlation coefficients for the individual replicates ranged from 0. Boxplot showing the numbers of quantified proteins of six squamous cell carcinoma tumour samples that were analysed on both an Orbitrap Fusion and a Q Exactive HF mass spectrometer both Thermo Fisher.
Technical reproducibility on different mass spectrometry platforms by direct comparison of the protein expression profiles of six squamous cell carcinoma tumour samples that were analysed on both an Orbitrap Fusion and a Q Exactive HF mass spectrometer both Thermo Fisher.
UBC Theses and Dissertations
As a first exploratory analysis, we performed a principal component analysis on the full proteomic data. To this end, we have investigated if any of the differentially expressed proteins were identified as genetic dependencies i. Volcano plot relating adjusted p values for differential protein expression to averaged normalised SILAC ratios of two replicates.
Highlighted in green are genetic dependencies that were also identified as differentially expressed proteins in our study. Shown are mean values and standard deviation. Statistical significance was assessed using Wilcoxon—Mann—Whitney test. Shown are mean and standard deviation; statistical significance was assessed using Wilcoxon—Mann—Whitney test.
Training was performed using a bootstrapping procedure to assess robustness of predictions and guard against overfitting. Using the full set of proteins as features in a crossvalidation approach, the predictor achieved a bootstrapped accuracy of We then assessed the extent to which reducing the size of the signature set would affect classification performance.
Using recursive feature elimination, we were able to reduce the set to about 1, proteins without affecting classification performance. Thereafter, the accuracy of the classification began to decline gradually, with a bootstrapped accuracy of This indicates clearly that reliable distinction between these closely related tumour types will only be possible if complex protein expression signatures, rather than single immunohistochemical markers, are used.
Survival analysis was used to assess classification performance on this final dataset. Classification accuracy using a linear support vector machine. The figure shows the bootstrapped accuracy achieved over the number of proteins included in the computation.
Feature selection was performed using recursive feature elimination. In the next step, and as an independent validation of our diagnostic proteomic signature, we analysed the protein expression profiles of 38 independent tumour samples, of which 19 were either confirmed primary SQCLC or primary HNSCC cases.
Similar to the crossvalidation analysis, the diagnostic accuracy in this independent validation cohort was Classification results of the clinical score left and the SVM prediction model right for the 51 lung tumours of unknown origin.
The p53 mutational status is shown. Differentiation between these two conditions is of fundamental clinical importance for therapeutic stratification. Interestingly, intersection of our proteomic data with functional genomic data revealed that some of the differentially expressed marker proteins were genetic dependencies in either of the two tumour types. Our data show that a distinction with clinically actionable accuracy requires a diagnostic signature rather than single markers.
Lois de lenz pdf viewer
The minimum number of features required to achieve this accuracy was 1, proteins. Importantly, in contrast to all established clinical criteria used for this purpose, our proteomic signature was able to stratify patients with lung tumours of doubtful origin into risk groups with significantly different prognosis. This finding underscores the assertion that a reliable distinction between such closely related tumour types is possible if multiparametric protein expression signatures are used.
None of the patients had received neoadjuvant chemotherapy. Informed consent was obtained from all patients. All procedures were conducted in accordance with the Declaration of Helsinki and institutional, state and federal guidelines. Tissue samples were analysed by light microscopy after counterstaining with Meyer's haematoxylin.
Loi de Lenz
PCR products from the same patient were pooled after treatment with FuPa reagent. For cell lysis, cells were washed with PBS and lysed with the same lysis buffer as described above.
Maisch GmbH. Solvent A consisted of 0.
All scan events were automatically parallelised. Other settings were the same as the ones on the Orbitrap Fusion. Cysteine carbamidomethylation was defined as a fixed modification. Minimum peptide length was set to seven amino acids, with a maximum of two missed cleavages.
At least two ratio counts were required for peptide quantification. Data analysis was performed in R version 3. Due to the agnostic approach pursued, no prior sample size computation was performed; all available samples were included in the analysis.
P values were corrected using the false discovery rate. Multivariate analysis was performed using the R package caret Kuhn, Feature selection was done using recursive feature elimination. In all analyses, the investigator was blinded for tumour types in the validation cohort and for patient outcome for the classification of tumours of unknown origin.
HB and TO conceived and supervised the project. LK performed bioinformatics and statistical analyses. FC and NVD performed bioinformatic analyses.
In the context of HNSCC, differentiation between lung metastasis and secondary squamous cell lung cancer has important prognostic and therapeutic implications. However, reliable biomarkers for differentiation between these tumour types are lacking because of their shared morphologies and genomic features.
National Center for Biotechnology Information , U.
Published online Aug Author information Article notes Copyright and License information Disclaimer. Thomas Oellerich, Email: ed. Corresponding author. Published under the terms of the CC BY 4.
This article has been cited by other articles in PMC. Associated Data Supplementary Materials. Table 1 Patient characteristics training cohort. Open in a separate window. Figure 1.
Comparative proteomics reveals a diagnostic signature for pulmonary head‐and‐neck cancer metastasis
Figure 2. Figure EV1. Technical reproducibility A. Figure 3.
Figure EV2. Figure EV3. Table 2 Patient characteristics immunohistochemistry cohort. Figure 4. Figure EV4.
Diagnostic signature for lung tumours of unknown origin A. Table 3 Patient characteristics validation cohorts. Statistical data analysis Data analysis was performed in R version 3.
Author contributions HB and TO conceived and supervised the project. Conflict of interest The authors declare that they have no conflict of interest. The paper explained. Supporting information.
Dataset EV1 Click here for additional data file. Dataset EV2 Click here for additional data file.
Dataset EV3 Click here for additional data file. Dataset EV4 Click here for additional data file. Dataset EV5 Click here for additional data file. Dataset EV6 Click here for additional data file. Review Process File Click here for additional data file. Cell Rep 23 : —