次世代定序資料模擬軟體的比較

Abstract

次世代定序技術可以在短時間內，產生大量的定序資料，因此科學家們可以以低廉的價格進行基礎研究，也更能探究基因體、轉錄因子和分子間交互作用等知識。此外，由於大量的資料產生，生物資訊學家及統計學家必須找尋新的方法來處理及分析資料。最近有很多的次世代定序模擬軟體已經被發表，假如這些模擬軟體能夠產生出與真實定序資料相似的模擬資料，將有助於進行方法的推導，以及實驗流程的規劃。本篇論文中，進行ART, FlowSim, MetaSim, SimSeq, 及wgsim等五種模擬軟體在模擬Roche 454 和 Illumina 定序資料與真實資料的比較。我們選擇了大腸桿菌以及水稻作為模擬的材料，透過序列組裝及序列比對的過程，得知在模擬Roche 454資料時，FlowSim需要花最長的時間進行模擬，而ART的模擬資料，比較接進真實資料的表現。另一方面，在模擬Illumina資料時，SimSeq所花的模擬時間為最長。當進行模擬大腸桿菌這類基因體較小的資料時，所有的模擬軟體在序列組裝及序列比對上的表現皆與真實資料相似，而在像是水稻這種較大的基因體時，所有模擬軟體除了ART外，在序列組裝上的N50 及最大contig長度，皆超出真實資料的長度。本研究只做了初步的序列組裝和序列比對的分析，並未考慮序列品質分數等其他因素，因此無法很明確的評斷模擬軟體的好壞，以後還需要多方面的研究，了解基因體的特性，才能決定何種模擬軟體較合適。

Next-generation sequencing technologies can sequence large amounts of bases in a short time, enhancing the fundamental biological research. Scientists could comprehend the knowledge about genomes, transcriptomes and interactomes by sequencing at low cost. In addition, because of the massive data generating by NGS, bioinformaticians and statisticians have to find new methods to process and analyze data. Many NGS data simulators have been proposed recently. If the simulator can produce data that are reasonably similar to the real data, it will help the inference about adequate methods and the setting of experimental workflow. In this thesis, we had compared five simulators, including ART, FlowSim, MetaSim, SimSeq, and wgsim, in application of simulating Roche 454 and Illumina platform data for E. coli and rice (Oryza sativa) genomes. The simulated data were compared with public-available real sequencing data through assembling and mapping to reference genome. For simulating Roche 454 data, FlowSim took the longest time to simulate; the computing time for other simulators are competitively shorter. ART generated data that were the most similar with the real data if comparing the results of assembly and alignment. While simulating Illumina sequencing data, SimSeq spent the most time on simulations. For simulations of small genome size Illumina date like E. coli, all simulators well illustrate the real results of assembly and alignment. However, while simulating lager genome size like rice, all simulators, except of ART, got over optimistic results in estimating the N50 and maximum contig length. In this thesis, we simply analyze data roughly by assembly and alignment, which is not enough to judge the pros and cons of simulators. Therefore, further research is needed and realize the characteristics of genomes to select proper simulators.