應用於微流體生物晶片之樣本製備技術

Abstract

作為近年生醫電子領域的新寵兒，實驗室生物晶片技術能於數公分見方的晶片上，實現過去需於實驗室方能完成之生化分析。而樣本製備程序(sample preparation)為各種生化分析中不可或缺的步驟，其品質顯著影響分析過程中的反應物成本。本論文將針對新一代實驗室晶片──微流體生物晶片上之樣本製備技術做一完整討論。 生物晶片上的樣本製備程序有三個主要的考量因素：混合模式、目標濃度數目、及參與的反應物種類。微流體生物晶片依其流體驅動方式主要可分為微流體生物晶片(digital microfluidic biochip, DMFB)及流式微流體生物晶片(flow-based microfluidic biochip)兩類，而不同的晶片架構適用不同的混合模式(mixing model)。其上的樣本製備程序，依所需要的目標濃度數目及參與的反應物種類，可再細分為單一/多重目標樣本濃度製備(single-/multi-target sample preparation)、以及雙/多反應物製備(two-/many-reactant sample preparation)等。 本論文第一部分(第二章、第三章)，針對當前最有潛力的數位微流體晶片，提出反應物用量最佳化樣本製備演算法。實驗結果證明，於單一目標樣本製備程序，我們所開發之演算法較現有方法能節省32%至52%不等的反應物成本。此方法更可進一步擴充，使其適用於多重目標濃度之樣本製備。與舊有方法相比，於一般狀況下(目標數為10)，能減少48%廢液量及37%反應時間。若需要製備的樣本數目更多，甚至能達到97%和73%的改善幅度。 論文第二部分(第四章)同樣基於數位微流體晶片架構，但討論主軸為多反應物樣本製備程序。由於其他反應物的加入，使樣本製備問題的複雜度更加提高。當前大部分舊有方法均無法處理多反應物樣本製備問題。為了降低問題複雜度，我們設計出基於配方矩陣(recipe matrix)的濃度表示法，以反映溶液內的不同反應物含量。並利用配方矩陣上的矩型合併(rectangle merge)操作，開發出能減少反應物成本的稀釋動作共用演算法。和當前可進行多反應物製備的方法相比，本文所述演算法平均能節省27%的反應物成本。 最後，本論文第三部分(第五章)討論現行應用較廣泛的流式微流體生物晶片。由於架構上的差異，流式微流體晶片能較前述之數位微流體晶片支援更多的混合模式，進一步提高了廢液再利用的機會。然而，目前尚未有任何針對流式生物晶片開發的樣本製備程序。故本文提出第一個針對流式微流體晶片的樣本製備演算法，藉由在製備過程中選用適當的混合模式，可將既有演算法產生之製備流程針對流式微流體晶片進一步最佳化。此方法可與所有輸出結果為二元樹之樣本製備演算法搭配，於搭配不同演算法的情形下，根據實驗結果可進一步降低37%至69％不等的反應物用量。

Lab-on-a-chip, one of the emerging promising technologies in bio-electronics, which can perform biochemical assays usually conducted in laboratories on a small device with only few square centimeters in size. In most biochemical assays, sample preparation is an essential step, and its quality significantly determines the whole assay cost. As a result, this dissertation mainly addresses the optimization techniques for sample preparation on a latest type of lab-on-a-chip – microfluidic biochips. There are three key factors in the sample preparation process: mixing model, number of target concentrations, and amount of reactants involved. Two microfluidic biochips are commonly utilized into applications; they are digital microfluidic biochips (DMFBs) and flow-based microfluidic biochips (FMFBs). The two kinds of chips have different fluidic driving mechanisms, and thus lead to different sets of available mixing models. Besides, in terms of the number of target concentrations as well as the reactants involved, sample preparation can also be classified as single-/multi-target sample preparation as well as two-/many-reactant sample preparation. In Part I of this dissertation (Chapter 2 and 3), a reactant minimization algorithms is presented to tackle the sample preparation problem on digital microfluidic biochips. According to the experimental results of single-target sample preparation, our algorithm can save 32% to 52% of reactant cost compared with previous arts. This algorithm can be further extended for multi-target sample preparation. If the number of target concentrations equals to 10, our algorithm can further reduce 48% of waste production and 37% of preparation time as compared with the existing algorithm. This reduction can be up to 97% and 73% if the number of target concentrations goes even higher. Part II of this dissertation (Chapter 4) focuses on the sample preparation problem with many reactants involved. The difficulty of sample preparation problem is strongly proportional to the number of reactants. So far, few algorithms can tackle the many-reactant preparation issue. To solve this problem, we develop a new concentration expression to present the contents of different reactants within a solution, called recipe matrix. Then, we also propose a common dilution operation sharing algorithm based on the rectangle merging operation on a recipe matrix for reactant minimization. Compared with the latest existing algorithm for many-reactant preparation, our algorithm can save up to 27% of reactant consumption on average. Finally, sample preparation on flow-based microfluidic biochip is discussed in Part III (Chapter 5). A flow-based microfluidic biochip can provide more mixing models than a digital microfluidic biochip according to its inherent structure, and thus can enable better ways to reuse waste solutions. However, no sample preparation algorithm for flow-based microfluidic biochip has been proposed yet. As a result, this dissertation presents the first algorithm for flow-based microfluidic biochips. It can further optimize an initial process produced by existing algorithms through re-selecting a better mixing model for every dilution operation during sample preparation. Our method is compatible with an existing sample preparation algorithm as long as its output is a binary tree. According to the experimental results, the proposed algorithm can save up to 37% to 69% of reactant amount.