於數位微流體生物晶片中可將反應物最小化 之多目標濃度及多反應物樣本製備技術

Abstract

樣本製備程序(sample preparation)為各種生化反應中不可或缺的步驟之一。多種原始生物樣本或反應試劑必須在此程序中進行稀釋或混合，以得到一個或一個以上所需的目標濃度(target concentration)。由於這些樣本或是試劑對整個生化反應的成本有很大的影響，因此應儘量減少試劑的成本。本篇論文將提出一個有效重覆利用相同濃度之中介液珠的樣本製備程序演算法。我們的演算法的主要概念為，將給定之多組目標濃度以一成分立方(recipe cube)表示，然後在成分立方上尋找配對(pair)，將他們組成配對集合(pair set)，這些配對集合暗示著減少反應試劑或是樣本用量的機會，各個配對集合會根據他們對反應造成的效益以及成本計算出分數，分數最高的配對集合會被挑選出來降低試劑成本，演算法會根據選出的配對集合來合併液珠並更新成分立方，此步驟會被執行直到無法在成分立方中找到分數高於零的配對集合。實驗結果顯示我們的演算法能夠減少當前最好的演算法21.3%的反應試劑成本。

Sample preparation is an essential process for most on-chip biochemical applications. During this process, different reactants are mixed to produce one or more than one specific concentration values. Since the cost of bioassay is directly influenced by the reactants, the reactant cost should be minimized whenever possible. Hence, in this thesis, we propose a sample preparation algorithm that is able to reuse the intermediate droplets with the same concentration value during the process. First, our algorithm represents the given target concentrations as a recipe cube. Second, it identifies all pairs in the recipe cube. The pair set, which consists one or more pairs, indicates the opportunities for reusing the intermediate solutions. Each pair set is assigned a score according to its gain and cost and the one with the highest score will be chosen as a candidate for our algorithm. Then the chosen pair is merged and the recipe cube is updated. Finally, the iteration terminates while there is no pair sets with positive score. Experimental results show that our algorithm can save reactant cost by up to 21.3% against the state-of-the-art method.