Need Help?
Copied to clipboard!

Disentangling the Effect of Black/African Ancestry from Confounding Variables on Colorectal Cancer Biology

This study examined the effect of African ancestry on AI assisted computation of tumor infiltrating lymphocytes (TIL%) scores from hematoxylin and eosin stained whole slide images (WSI) assembled from three cohorts: 1.) African vs. European ancestry WSI downloaded from the TCGA-COAD-READ database; 2.) African vs. European ancestry WSI assembled from three US medical centers; 3.) Nigerian WSI assembled from a single Nigerian medical center. There were 33 African ancestry WSI suitable for analysis in the TCGA cohort and 48 African ancestry WSI in the independent US cohort. In addition to self-described African ancestry, the additional independent covariables were: 1.) Mismatch repair enzyme-deficient/ microsatellite instability-high vs. MMR-proficient/MSI-Low +microsatellite stable (MMR-d/MSI-H vs. MMR-p/MSI-L+MSS); 2.) Categorical age at diagnosis/resection (<50 y, 50-75 y, >75 y); 3.) Sex (Male vs. Female); tumor location (Right vs. Left + Rectum); tumor stage (I+II vs. III+IV), tumor grade [G1+ G1/G2+G2 vs. G2/G3 +G3+mucinous (regardless of G)]. For the 242 TCGA cohort there were 241 parallel whole transcriptome data generated from parallel frozen tumor samples. For the independent US cohort there were 87 parallel targeted whole exome transcriptome data generated from the same formalin fixed paraffin embedded (FFPE) block. The TCGA and independent US cohort computed TIL% scores were correlated separately with CIBERSORTx derived total lymphocyte and T-cell abundance scores, and with CXCL10 and CCL5 gene expression values. mRNA gene expression count data (from STAR processing of the RNA-sequence data) and WSI for the subjects used in this study are available from the Genome Data Commons (GDC).