# Non-coding genetic variants underlying higher prostate cancer risk in men of African ancestry

## TL;DR

This paper identifies non-coding genetic variants linked to higher prostate cancer risk in men of African ancestry, improving risk prediction.

## Contribution

The study introduces non-coding SNPs affecting prostate enhancer function, offering novel insights into prostate cancer risk in African ancestry populations.

## Key findings

- Identified ~2400 non-coding SNPs with higher frequency in African ancestry men that may influence prostate cancer risk.
- These SNPs modulate transcription factors like FOX, HOX, and AR, affecting cancer-related pathways.
- Incorporating these SNPs improves polygenic risk scores for prostate cancer in African ancestry populations.

## Abstract

Prostate cancer (PrCa) incidence and severity vary across ancestries; men of African ancestry (AA) are more likely to be diagnosed and die from PrCa than those of European ancestry (EA). Current polygenic risk scores, even from multi-ancestry GWAS, do not fully capture population-specific genetic mechanisms, especially those mediated by non-coding regulatory single nucleotide polymorphisms (SNPs). Using a deep learning model of prostate enhancers, we identify ~ 2000 SNPs, potentially affecting enhancer function, with higher alternate allele frequency in AA men, that may affect PrCa risk. These SNPs may promote cancer via two mechanisms: increased enhancer activity leading to immune suppression and telomere elongation or decreased activity causing de-differentiation and apoptosis inhibition. Identified SNPs predominantly modulate binding of key transcription factors such as FOX, HOX, and AR – the first was experimentally validated. Incorporating these SNPs into a polygenic risk score improves PrCa risk assessment beyond existing GWAS-identified variants.

Current polygenic risk scores for prostate cancer do not leverage biological mechanisms and remain inadequate for patients with African ancestry. Here, the authors employ a deep learning model to identify 2,407 non-coding polymorphisms with greater frequency in African American individuals that may affect enhancer activity in prostate cancer-related pathways, leading to more accurate polygenic risk scores.