隨機灘設計與反物模式

Abstract

一個 clone 是一小段 DNA 序列, clone library是儲存了大量 clones 的地方。從 clone library 裡找出擁有特定性質的 clones, 我們稱之為正物, 便是 clone library 的檢測問題。由於 clone library 內存有數量非常龐大的 clones, 逐一檢測非常的浪費時間與金錢。將一群 clones 放在一起, 並稱此集合為一個攤(pool) 。針對一個攤做檢測時, 若此攤中含有正物, 我們便會得到正反應, 反之則得到負反應。我們用檢測一個攤來取代逐一測試所有 clone, 此外, 我們還同時檢測好幾個不同的攤, 藉此節省測試次數與時間。攤設計便是數個被同時檢測的攤, 我們利用它來有效地找出正物。 由於攤的製造費時且過程相當繁複, 我們必須使用簡單的攤設計以便讓製作過程自動化。然而簡單的攤設計通常就代表了, 它並不足以幫我們找出所有的正物。於是一個攤設計是否能讓我們找出夠多的正物, 便成了檢驗它們是否適用的標準。隨機攤設計便因為兼具了構造簡單以及成效良好的優點, 越來越受到重視。本論文提供了現有的五種隨機攤設計的成效分析。 此外, 由於DNA分子間的作用, 有時正物會與某些亦存在於 clone library 裡的 clones, 我們稱之為反物, 產生化學反應。若是受測試的攤中, 同時包含了正物與反物, 我們會誤認為此攤中並不含有正物。本論文提供了一類攤設計, 在即使有反物與實驗誤差的情況下, 也能有效率地找出所有正物。

A clone is a DNA subsequence and a clone library is a large collection of clones. Screening of a clone library is to identify all specific clones, we call them positives, in the clone library. Because the number of clones in a clone library is very large, check each individual clone one by one is a waste of time and money. So instead, we choose a set of clones to form a pool and screen each pool as a unit. If there is a positive in the pool, we get a positive result, otherwise, we get a negative result. Further, we screen the pools in parallel to save time. Such a set of pools is called a pooling design which helps us solve the clone library screening problem efficiently and economically. Because the construction of a screening environment, i.e. a pool, is time consuming and complicated, we need simple pooling designs to facilitate the automation of pools construction. However, simple designs usually are not powerful enough to identify all positives. Hence whether a pooling design can identify as many positives as we need becomes a measure of its performance. With simple structure and good performance, random pooling designs get more and more attention recently. This thesis provides the performance analysis and comparison of five available random pooling designs. Further, due to the interaction between DNA molecules, positives might interact with some clones, called inhibitors, to lose their positiveness. This thesis provides a class of pooling designs which identify all positives efficiently even in the presence of inhibitors and experiment errors.