利用結構分解與抽象處理搜尋核糖核酸結構

Abstract

近年來，在核醣核酸的相關研究上有許多新發現，研究發現核醣核酸有許多我們以往不清楚的功能，而這些功能在生物學上皆扮演著重要的角色，2006/10/02公佈的諾貝爾醫學獎得獎的主題「RNAi (RNA interference, RNA干擾現象 )」就是最好的例子。由於核醣核酸的功能與其二級結構有密切的關係，因此若能提供核醣核酸二級結構的相關資訊給生物學家，則能協助他們加快檢驗出核醣核酸的功能。在本研究中，我們提供一個生物資訊的方法，自一群相關的核醣核酸序列中，找出其二級結構的共同結構元。我們以現有的單一核醣核酸二級結構預測系統作為前處理器，預測出每條序列數個可能的二級結構，將其結果使用我們所設計的演算法做分解並抽象化之後，使用Gibbs-like流程方法來找出其共同結構元，並且為了能提升系統的效能，設計了一些新的對核醣核酸結構做分析的演算法，它們的時間復雜度都比現今提出的演自法來得更有效率，使得我們的系統能在記憶體和執行時間上遠勝於其它的系統。為了驗證我們系統的準確定與效能，我們從Rfam中下載數個RNA家族的資料來做測試，實驗結果也顯示出本方法有不錯的表現。

Motivation: RNA molecules are the key players in the biochemistry of the cell, playing many important roles in regulation, catalysis and structural support. Many functional RNAs have evolutionarily conserved secondary structures in order to fulfill their roles in a cell. Although current approaches can identify common structure motifs from a set of RNAs, they typically rely on the assumption that the given sequences are from a single family, which is not necessarily true in practice. Results: We develop a new method based on structure decomposition and Gibbs sampling to predict consensus structure motifs in unaligned RNA sequences. Unlike most current approaches, our method is applicable to a set of mixed sequences from different families, and is able to predict multiple motifs for multiple families. Furthermore, as we separate motif finding from sequence folding in our system, new folding algorithms other than Mfold or RNAfold, etc. can be easily integrated with our motif finding process. Extensive testing on 17 families from Rfam shows that our method competes well with other current tools in single family predictions. As for multi-family predictions, experiments also demonstrate that our new approach outperforms recent alternative methods.