利用英國及台灣人體資料庫探討跨種族身高相關研究

Development of a prediction model based on genetic variants for height and associated phenotypes by using Taiwan and UK biobanks

鄭毓嫻(Yu-Hsien Cheng)

指導教授 ： 盧子彬

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2021年)

https://doi.org/10.6342/NTU202101335

摘要

研究背景: 身高是一種經典的多基因性狀，受到多個基因控制，這些影響身高的基因在不同種族中可能有所不同。身高是人類可遺傳的複雜性狀之一，其遺傳力估計在80％左右。先前的流行病學研究已經證實，身高與多種疾病（包括癌症，中風和心房顫動）的風險有關，身高具有高度的遺傳力和種族異質性。雖然身高是一個非常容易觀察的特徵，歐洲族群與亞洲族群的平均身高差異也顯而易見，但是目前缺乏跨種族的研究。此外本研究還想探討身高的多基因風險評分與疾病的風險之間的關聯。
材料與方法: 使用台灣生物資料庫(Taiwan Biobank)，並納入14,766名參與者和557,752個位點；英國人體資料庫包括424,104個參與者和456,028個位點。為了找出分別影響歐洲族群和亞洲族群身高的遺傳變異，分別在台灣人體資料庫和英國人體資料庫中進行全基因組關聯研究（GWAS）。最後把影響身高相關的位點透過多基因風險評分(polygenic risk score)建模並且分別預測台灣人的身高和歐洲人的身高，接著做一個跨種族比較，深入探討變異位點跟疾病之間的相關性。
結果: 透過GWAS分析結果，發現不同人種間存在著不太一樣的變異位點。在台灣族群中，多基因風險評分總體上解釋了35％的身高; 在調整了性別之後，解釋力增加到了85％。 另一方面，在歐洲族群中，多基因風險評分總體上解釋了約16％的身高;調整性別後，解釋力增加到約65％。
透過交叉比較，我們以歐洲族群為訓練集，以台灣族群為測試集(使用英國族群的權重預測台灣族群)。多基因風險評分僅解釋了身高的2.6％。調整性別後，解釋力也才增加到50％。如我們預期，當群體間遺傳效應的相關性降低時，預測準確性會下降。 另一方面，我們將台灣族群作為訓練集，將歐洲族群作為測試集(使用台灣族群的權重預測歐洲族群)。同樣地我們發現預測能力也是下降，多基因風險評分僅解釋了不到身高的1％，可以預期在調整性別後，預測力也只有剩下約50％。此外在多基因風險評分與疾病的風險之間的關聯，在乳癌、消化器官類的癌症都看到一個正向的關聯。
結論: 透過多基因風險評分進行跨種族間身高的比較，可發現不同種族之間影響身高的位點不太一致。在使用多基因風險評分來預測身高，此模型的預測力與過往研究的結果一致。而在疾病關聯性方面，發現隨著身高越高，罹患乳癌、消化器官類的癌症的風險也有越高的趨勢。

關鍵字

身高 ； 多基因風險評分 ； 台灣生物資料庫 ； 英國生物資料庫

並列摘要

Background: Body height is a classical polygenic trait, controlled by multiple genes that may be diverse in different ethnic populations. Obviously, body height is one of the most heritable human complex traits, with its heritability being estimated between 80–90%. Previous epidemiological studies have confirmed that body height is associated with the risk of various diseases including cancer, stroke and atrial fibrillation, which has profound genetic disposition and shows high degree of ethnic heterogeneity. While body height is a very superficial trait, epidemiological studies have confirmed that there is an association between height and certain diseases. However, there is a lack of trans-ethnic study currently. In addition, explore the relationship between genetic score of height and odds ratios of certain diseases.
Methods and Materials: After quality control, Taiwan biobank includes 14,766 samples and 557,752 variants; UK biobank includes 424,104 samples and 456,028 variants. To find out genetic variants associated with body height between Europeans and Asians, genome-wide association studies (GWASs) will be conducted in Taiwan biobank and UK biobank. Polygenic risk score is used to do modeling and predict Taiwanese body height and European body height, respectively.
Results: Through the results of GWAS analysis, it is found that there are different mutation sites among different races. Polygenic risk score in aggregation explained about 35% of body height in Taiwanese population. After adjusted for gender, the squared correlation increased to 85%. On the other hand, in European population, PRS in aggregation explained about 16% of body height. After adjusted for gender, the squared correlation increased to 65%.
Through cross-validation, we use the European ethnic group as the training set and the Taiwan ethnic group as the test set (using the weight of the British ethnic group to predict the Taiwanese ethnic group). The polygenic risk score after adjusting the gender, the squared correlation has only increased to 50%. As we expected, when the prediction accuracy decreased further when genetic effects were less correlated across populations. On the other hand, we use the Taiwanese ethnic group as the training set and the European ethnic group as the test set (using the weight of the Taiwan ethnic group to predict the European ethnic group). It can be expected that after adjusting the gender, the squared correlation will only be left about 50%.
In addition, there is a positive association between the polygenic risk score and the risk of breast cancer and digestive organ cancer.
Conclusion: Through the polygenic risk score to compare the height between different ethnicities, the prediction accuracy decreased further when genetic effects were less correlated across populations. In using the polygenic risk score to predict body height, the squared correlation of this model is consistent with the results of previous studies. In terms of disease relevance, it was found that the higher the body height, the higher the risk of breast cancer and digestive organ cancer.

並列關鍵字

body height ； polygenic risk score ； Taiwan Biobank ； UK Biobank

參考文獻

1. Sinclair, D., Human growth after birth. 1973: Oxford University Press.

2. Maher, B., The case of the missing heritability: when scientists opened up the human genome, they expected to find the genetic components of common traits and diseases. But they were nowhere to be seen. Brendan Maher shines a light on six places where the missing loot could be stashed away. Nature, 2008. 456(7218): p. 18-22.

3. Kabat, G.C., et al., Adult height in relation to risk of cancer in a cohort of Canadian women. International journal of cancer, 2013. 132(5): p. 1125-1132.

4. Green, J., et al., Height and cancer incidence in the Million Women Study: prospective cohort, and meta-analysis of prospective studies of height and total cancer risk. The lancet oncology, 2011. 12(8): p. 785-794.

5. Shimizu, Y., et al., Adult height and body mass index in relation to risk of total stroke and its subtypes: the circulatory risk in communities study. Journal of Stroke and Cerebrovascular Diseases, 2014. 23(4): p. 667-674.