CyTOF是一種利用質譜方式藉由抗體與同位素純元素標定並且霧化樣品在飛過時間質譜儀時分析金屬信號在而達到實時的抗體表達的技術，可同時測定30-40種標記。一開始我們是使用手動篩選，並使用決策樹產生的標記作為參考方便進行以後資料分析。但因為這種挑選會根據研究人員的判斷不同而有所差異，而希望用自動的方式進行挑選。所以我們設計了一套從自動圈選到決策樹的分析流程。我們做的方法就是藉由數學運算的方式來找出每個檔案不同的門檻值，由不同的算法及參數調整以找出適合每個樣本的算法。再由決策樹將這些檔案算出其中可以用來分類細胞型態的重要標記，以及各類不同算法下的準確度差異，並且與其他人所做的方法做比較。我們所開發出來的挑選方法中可以從自動圈選調整多個參數到決策樹產生不同樹狀圖及標記。由這個方法可以提供給使用CyTOF的研究人員有更多種選擇去分析各種資料及產生適合不同實驗的結果;CyTOF is a technique that uses mass spectrometry to calibrate antibodies and isotope pure elements, and analyzes the metal signal of the atomized sample when flying through a time mass spectrometer to achieve real-time antibody expression. It can simultaneously measure 30-40 markers. At the beginning, we used manual screening and used the tags generated by the decision tree as a reference to facilitate future data analysis. But because this kind of selection will be different according to the judgment of the researchers, we hope to choose it in an automatic way. We designed a set of analysis processes from autogating to a decision tree. The method we do is to find different thresholds for each file by means of mathematical operations, and adjust the different algorithms and parameters to find the algorithm suitable for each sample. These files are then used by the decision tree to calculate the important calibrators that can be used to classify cell types, as well as the differences in accuracy under various different algorithms, and compare with the methods made by others. The selection method we developed can adjust multiple parameters from autogating to decision tree to generate different dendrograms and labels. This method can provide researchers using CyTOF with more options to analyze various data and produce results suitable for different experiments.
