請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/8306
標題: | 透過多批次外表型收集策略選拔優良基因型 A Sequential Batch Phenotyping Strategy for Detecting Superior Genotypes |
作者: | Zhen-Yu Tu 杜鎮宇 |
指導教授: | 廖振鐸(Chen-Tuo Liao) |
關鍵字: | 多批次外表型收集策略,多性狀,選拔指標,r-score,GBLUP, Sequential phenotyping strategy,Multiple traits,Composite selection index,r-score,GBLUP, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 使用多批次外表型收集策略來從一組候選族群中找到一群優良基因型群體,以達到能節省外表型資料的蒐集,進而找到優良基因型群體。在本研究中我們假設候選族群已擁有基因型資料,並分多次選取部分個體收集其外表型,而後使用具有外表型資料的個體建立GBLUP多性狀模型、並估計候選族群個體的基因型值、對於不同性狀給予不同權重後相加成一個選拔指標,並進行排序。其中用於選取訓練族群個體的方法有r-score、M-PGV、EI-PGV以及EI-PGV-fwd,而所有方法的第一組起始個體選取皆使用r-score的方法,因為r-score只需要使用基因型的資訊而不需要考慮外表型的資訊。多性狀模型的應用讓我們同時針對多個性狀進行估計、然後根據不同性狀的重要性進行加權總合、最終得到的值稱作composite selection index (CSI)。針對排序後的CSI則使用correctly identified proportion (CIP)以及normalized discounted cumulative gain (NDCG) 作為評估指標,這兩項指標可以對感興趣的前幾名個體進行評量,且NDCG還多考慮了排序的正確性。經由上述的流程,最終能夠輔助我們選拔出個體來進行外表型資料蒐集、使得模型有良好的估計與排序,進而有效率的找到優良的基因型群體。 A sequential phenotyping strategy is proposed to detect a set of superior genotypes efficiently from a candidate population. In this study, we assume that all of the individuals in the candidate population have been already genotyped. The iterative searching process is composed of the following steps. Step 0: a starting training set is determined from the candidate population according to the r-score algorithm. Step 1: a multiple-trait GBLUP model is trained using the phenotype and genotype data of the current training set. Step 2: a composite selection index (CSI) is constructed and estimated for each individual in the candidate population with genotypes based on the resulting multiple-trait GBLUP model. Step 3: two assessment indices, correctly identified proportion (CIP) and normalized discounted cumulative gain (NDCG) are calculated based on the estimates of CSI for a set of candidate individuals, and are used to evaluate the accuracy for the detection of the superior individuals. Step 4: four acquisition functions, r-score, M-PGV, EI-PGV and EI-PGV-fwd, are used to select additional training set added with the current training set. We further provide a stopping rule for the sequential strategy for practical applications. Three genome datasets are analyzed to illustrate our proposed sequential phenotyping strategy. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/8306 |
DOI: | 10.6342/NTU202002551 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 農藝學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-0608202015295500.pdf | 2.93 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。