本文針對不同測量方式所產生的基因表現值進行兩序列間的相似性評估與整合。Aach和Church（2001）提出兩序列時間對應點之演算法，利用兩序列表現值之距離矩陣進行動態規劃法找出兩序列的時間對應點路徑；但在兩序列尺度不一致時，其結果並不理想。本文將兩序列表現值排序並產生排序差矩陣，以此矩陣進行動態規劃法，以改善尺度所造成的問題。此外，我們亦利用面積(或面積比列)的概念來評估兩序列的相似性，並找出面積比例的分佈及提出進行整合之準則。最後，若序列資料通過整合準則，我們可建立常態統計量以檢驗此兩表現值序列進行整合之效果。

We evaluate the similarity of two gene expression sequences from different experiments, and if possible, merging the two sequences. Aach and Church ( 2001 ) developed simple time warping algorithm to align gene expression time series, using distance matrix of two sequences through dynamic time warping to find the time warping path. When the scale of the two sequences are different, however, the results are doubtful. We sort the sequences in order, and use rank difference instead of Euclidean distance to find the optimal path. We use area ratio to evaluate the similarity of two gene expression sequences. Simulation is conducted to find a reasonable threshold in merging two sequences. Finally, we build one statistic to evaluate the performance of merging the two sequences once they pass through the threshold.

1. John Aach and George M. Church (2001), " Aligning gene expression time series with time warping algorithms " , BIOINFORMATICS, Vol. 17 no. 6 2001, Pages 495–508.