Based on the publication “Whole transcriptome profiling of patient-derived xenograft models as a tool to identify both tumor and stromal specific biomarkers,” James R. Bradford et. al.; DOI: 10.18632/oncotarget.8014. We have a total of 79 PDX (patient-derived xenograft or immuno-compromised mice with human tumors) samples: 37 lung and 19 breast cancer samples. We wanted to study the crosstalk between tumor and stroma cells using differences between tumor and stromal “transcriptomes.” By eliminating the noise, we can identify tumor- and stroma-specific biomarkers for subtypes of lung and breast cancers.
Next Generation Sequencing data was taken from 79 PDX models representing multiple cancer types (19 x breast, 37 x lung, 8 x colorectal, 7 x ovarian, 3 x endometrial, 2 x pancreatic, 2 x ampullary, 1 x leukaemia). http://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-3980/
- Mapping using TopHat (species-specific mapping)
- Detection of Isoforms using Cufflinks
- Creating a GTF file of isoforms using Cuffmerge
- Mapping transcripts on the transcriptome using Bowtie-2t
- Normalization and post-processing of data
- Previewing Results with PCA: looking for artifacts
- Batch Effect Correction
- Cross-talk Association: Clustering of Subsets (BiAssociation Approach)
Presence of a batch effect can be seen on the second principal component in the Principal Component Analysis (PCA) of genome-wide transcriptomic profiles:
For the identification of tumor-stroma crosstalk the BiAssociation method was applied. Expression profiles of different sets of genes can give different clustering of the same set of samples. BiAssociation identified and bi-associated a group of tumor genes and a group of stromal genes which led to identical or extremely similar clustering results. Three pairs of bi-associated groups of genes with the highest association score were analyzed. All pairs led to biologically meaningful results.
- Differentiation of small cell lung carcinoma samples (clusters of tumor and stromal genes)
- Lymphomagenesis samples that all had high levels of B-cell markers and two of these samples correspond to solid tumors are likely explained by the presence of Epstein-Barr virus.
- Stromal-specific samples, these samples had significantly higher expression of stromal genes then the other samples.