基因預測軟體在水稻白葉枯病抗性基因的應用性評估及分析平台建立

Abstract

利用基因預測軟體在含有水稻白葉枯病抗性基因(Xa)結構特徵的大片段序列集合中，找出最有可能的Xa 候選基因，再以基因定位試驗進行鑑別，可加速水稻抗病基因之發現。為找出適用於目前已知序列之Xa 基因結構型態的基因預測軟體，以利未來之Xa 基因預測工作，本研究蒐集以馬可夫模式為基礎所開發的5 種基因預測軟體：GlimmerM、GlimmerR、RiceHMM、GeneMark.hmm 及GENSCAN，以Xa1、Xa21、Xa26 及Xa27四個基因已知的18 條DNA 序列為測試資料，分別在核苷酸與外顯子之層次上評估其預測的靈敏性和專一性。分析結果顯示在核苷酸與外顯子兩個層次上，皆以GeneMark.hmm 軟體之準確度最高，其次為GENSCAN 軟體。另外，為使基因預測結果之展現及比較能更一目了然，亦以Bioperl模組程式設計出一套基因模式化視覺平台，提供將所預測之序列結構以不同顏色圖形瀏覽。據此建立了適用於水稻白葉枯病抗性基因預測的簡易分析程序與可應用分析平台。 To accelerate the gene finding of rice resistant to disease, it is essential to use gene prediction programs to find the most probable candidate, rice bacterial blight resistance gene (Xa), encoded in large rice sequence sets with Xa gene structure, and then perform gene mapping toexperimentally verify the prediction result. We believe a more realistic evaluation of various gene prediction tools on currently data sets with Xa gene structure would help our future work in Xa gene prediction. Thus, we collected five gene prediction programs based on Markov models, i.e., GlimmerM, GlimmerR, RiceHMM, GeneMark.hmm and GENSCAN, and used a total of 18 complete rice DNA sequences of bacterial blight resistance genes, i.e., Xa1, Xa21, Xa26 and Xa27, as test data. Sensitivity and specificity measures were employed to evaluate their gene prediction accuracy at both nucleotide and exon levels. Results showed that GeneMark.hmm was the most accurate in both nucleotide and exon prediction, and followed by GlimmerM. We also used Bioperl modules to design a gene modeling visualization platform that enabled a clear comparative display of the sequence structure predicted by the five programs. The graphic output provided a number of pre-defined color schemes that assisted in visual identification of the gene prediction differences among the five prediction programs. A simple analytical procedure and an analytical platform were thus established for gene prediction in rice bacterial blight resistance genes.