| 摘要: | 背景與動機：研究顯示結合常見變異的多基因風險評分（Polygenic risk score, PRS）和罕見致病性之變異（Rare pathogenic mutations, RPMs）可以有效地評估前列腺癌的風險。本研究旨在建立一種遺傳基因風險預測模型，以更好地鑒別東亞人群中的惡性前列腺癌病人。
研究方法：使用台灣精準醫療計畫（Taiwan Precision Medicine Initiative, TPMI）的台中榮民總醫院參與者，共有1054名男性前列腺癌患者來建構台灣人群的前列腺癌多基因風險評分模型，並另外使用424例病患作為外部驗證。將病例組根據其癌症期別分成高風險組和低風險組，以評估並比較使用C+T方式建立的PRS的表現。此外，本研究還納入由台中榮民總醫院提供的1023名具有全外顯子基因定序(Whole exome sequencing, WES)的前列腺癌男性用於罕見變異分析，並使用Sequencing Kernel Association Test-Optimal (SKAT-O)的統計方法找出疾病相關顯著基因。最後，通過使用Individualized Coherent Absolute Risk Estimator (iCARE)絕對風險估計工具，本研究計算了在不同PRS五分位數的組別中，罕見變異的?帶者與非?帶者的前列腺癌終身絕對風險。
結果：利用42個單核?酸多態性（Single Nucleotide Polymorphism, SNP）點位建構的PRS模型來預測台灣人的前列腺癌風險，該模型在高風險組表現的ORperSD=1.75(95% CI, 1.45-2.11)和AUC=0.658, 而在低風險組中的ORperSD=1.35(95% CI, 1.13-1.62)和AUC=0.591。當使用40-60%PRS分位數組別作為參考組時，在80-100%PRS分位數的高風險組病人的風險增加了3.86倍。根據SKAT-O在罕見變異分析的結果，我們發現共三個基因：ATM, BARD1和AR有達到統計上的顯著差異。將這些基因與PRS結合，我們可以估計在80-100%PRS分位數的罕見變異?帶者的前列腺癌終生絕對風險可達到46.89%，約為80-100%PRS分位數中非?帶者終生絕對風險（25.21%）的兩倍。
結論：本研究利用台灣族群所開發及驗證的PRS模型，在預測惡性前列腺癌方面與過去的相關研究有相當的表現。另外，將在東亞人群中發現的罕見變異整合到遺傳風險預測模型中，可以提高整體絕對風險估計的準確性。因此，結合PRS和RPMs的效果可以有效地輔助前列腺癌患者的風險分層評估。
BACKGROUND：
Recent studies have shown that combining the effect of common variant polygenic risk score (PRS) and rare pathogenic mutations (RPMs) could improve the risk assessment for male Prostate cancer (PCa). No currently validated genetic risk prediction tools capable of identifying and screening high-risk PCa, here we aim to develop a genetic risk prediction model to better discriminating aggresive PCa in East Asians.
MATERIALS AND METHODS：
A total of 1,054 microarray genotyped PCa patients recruited in Taiwan Precision Medicine Initiative (TPMI) at Taichung Veterans General Hospital, Taiwan (TCVGH) were included to generate an ethnic-specific PRS, with an additional 424 cases for external evaluation. The patients were categorized into high-risk and low-risk groups according to their cancer staging to assess the performance of PRS constructed using the clumping and p value thresholding (C+T) method. Furthermore, 1,023 PCa males with whole exome sequencing (WES) data were included in our study for rare-variant analysis. We performed the Sequencing Kernel Association Test-Optimal (SKAT-O) for genes discovery. Finally, by employing Individualized Coherent Absolute Risk Estimator (iCARE) tool, we computed the lifetime absolute risks of developing PCa for carrier versus non-carrier of RPMs in different PRS quintiles while accounting for competing risk of dying from causes apart from PCa.
RESULTS：
A 42-SNPs PRS model was created to predict Taiwanese PCa risk, showing an ORperSD of 1.75 (95% CI, 1.45-2.11) and AUC of 0.658 in the high-risk group, compared to ORperSD of 1.35 (95% CI, 1.13-1.62) and AUC of 0.591 in the low-risk group. We observed a 3.86-fold increased risk for high-risk PCa in the highest PRS quintile when utilizing the middle PRS quintile as the reference. Our gene-based SKAT-O test results revealed three statistically significant genes within the protein-truncated variants (PTVs): ATM, BARD1, and AR. Incorporating these genes with PRS, the absolute risks estimated of PCa for carriers in the top PRS quintile would attain risk level of 46.89% by age 90, which is approximately two-fold higher than non-carriers (25.21%) in the top PRS quintile.
CONCLUSIONS：
Our PRS for aggressive PCa, which developed and validated in a Taiwanese-based cohort, has shown comparable performance to relevant studies. Integrating rare variants discovered in the East Asian population into genetic risk prediction model can enhance overall absolute risk estimation. These findings suggest that the combined effect of PRS and RPMs could facilitate risk stratification in patients with PCa. |
