Introduction:

African American (AA) men with prostate cancer (PCa) often harbor more aggressive disease than those of European American (EA) descent, with higher overall incidence and PCa-related deaths. However, the impact of tumor biology on disparities in outcomes is currently controversial. Here, we performed gene expression analysis of PCa specimens from AA and EA men to characterize the biology of aggressive disease in these populations. 

Methods:

Whole-transcriptome sequencing of PCa specimens from 350 AA and 452 EA men who underwent radical prostatectomy was performed. Aggressive disease was defined as Gleason Grade group (GG) ≥3 and non-aggressive as GG <3. Differential Gene Expression (DEG) analysis via the limma R package, with correction for multiple comparisons using Benjamini-Hochberg with a false-discovery rate (FDR) of <0.05 and Random Forest (RF) classifier using the VSURF R package, were used to evaluate DEG between GG ≥3 versus GG <3. Two signatures were developed to predict aggressive PCa in AA and EA men by employing a penalized logistic regression model, fitted using a linear model with elastic-net regularization of glmnet R package. For model development, we used 67% of the AA and EA cohorts as training sets and 33% as validation sets. We used the pROC R package to construct a receiver operating characteristic (ROC) curve and calculate area under curve (AUC).

Results:

There were a total of 101 (29%) and 108 (24%) men with aggressive PCa in the AA and EA cohorts, respectively. In the AA cohort, 37 genes were differentially expressed (FDR<5%) between the GG ≥3 versus GG <3 groups. A 9-gene signature using an intersect of DEG and RF genes was developed to predict aggressive PCa in AA men, with AUC=0.74 in the training set and AUC=0.86 in the validation set. In the EA cohort, 23 genes were differentially expressed (FDR<5%) between aggressive PCa versus non-aggressive disease. A 7-gene signature was developed to predict aggressiveness with AUC=0.75 in the training set and AUC=0.72 in the validation set. There was no overlap of genes in the signatures developed for both cohorts (Figure 1). 

Conclusion:

The molecular profile of aggressive PCa in AA men is different from that of EA men. Future investigation to link these signatures to oncological outcomes is warranted to elucidate the impact of cancer biology on racial disparities in PCa. 

Funding: Michigan Prostate SPORE