肝臟與經修飾奈米碳管毒性模型

Abstract

近年來高通量篩選被廣泛應用在生物醫學及生物資訊領域上，傳統的生物實驗非常的耗費時間以及金錢。本篇論文是利用不同機器學習的方法建構了電腦分類模型，對資料做虛擬的高通量篩選，並針對兩個不同的主題進行探討和研究。 第一個主題是針對細胞色素P450超級家族中五種最常見酶的抑制的探討。細胞色素P450超級家族是在肝臟中涉及藥物代謝與生物激活作用的主要酶類，在本論文中探討的五種酶負責了人類體內90%的異質物與藥物的代謝，若這幾種主要代謝的酶被抑制，則會造成肝臟的傷害和毒性。希望能夠透過對這五種酶已知活性的小分子化合物，以機器學習的方法和不同特性的描述值歸納出對這五種酶的抑制最有預測力的數值模型，並和現有的方法進行比較。 第二個主題是奈米碳管活性的研究，近年來奈米碳管常常被應用在生物醫學上，當作生物感測器或是藥物攜帶者，所以奈米碳管對於人體是否有毒性越來越被人們所重視。在本論文當中，希望能夠透過一組經由奈米碳管的表面的修飾來降低在人體內所造成的毒性的實驗數據，經由基因演算法，找出針對牛血清白蛋白、碳酸酐酶、胰凝乳蛋白脢、血紅素以及細胞存活率、一氧化氮反應，這六種活性的最佳數值模型。其中前面四個標的是探討蛋白質結合的情況，後面兩個標的是探討細胞存活力及相關的人體免疫反應，最後進行模型上的歸納以及結構上的探討。

In recent years, high-throughput screening (HTS) is widely used in the field of biomedical and biological information. Bioassays are more time and money consuming when compared to a computational screening model. A toxicity prediction model can serve as a tool for interpretation of structural contribution to the end point or a pure virtual screening mathematical model. In this thesis, we constructed and discussed the use of computational machine learning method with hepatotoxicity and nanotoxicity endpoints to construct prediction models. Five of the most popular cytochrome P450 (CYP) superfamily enzymes in human liver inhibition models were constructed and discussed. CYPs are the major enzymes involved in drug metabolism and bioactivation. The five enzymes explored in this thesis account for 90% of the xenobiotic and drug metabolism in human body. Inhibition of CYP enzymes will lead to inductive or inhibitory failure of drug metabolism.1 By utilizing machine learning methods and different types of the descriptors, statistical models are built and further compared with the literature. In addition, a rule-based CYP inhibition prediction online server CypRules (http://cmdd.csie.org/~b96002/P450) is built based on the models reported in this study. CypRules can predict and provide structural ruleset for CYP inhibition. With the fast execution performance, it can be used for virtual high-throughput screening (VHTS) for a large set of testing compounds. In the second part of this thesis, we modify the surface of carbon nanotubes to find the relation between attached decorators and the complex toxic activity. There are six types of biological activities to be researched which are bovine serum albumin (BSA), carbonic anhydrase (CA), chymotrypsin (CT), hemoglobin, cell viability and NO response. The first four biological activities are related to protein binding, while cell viability and NO response are related to cell apoptosis and human immune response. Finally, we discuss results between models and structural features of compounds.