輔助人工智慧藥物開發之最佳化學分子相似性特徵衡量指標

Abstract

新藥研發廢日曠時，整個新藥開發的流程包括了早期候選藥物的探索與療效確立，以及產品開發之臨床前動物試驗與臨床人體試驗。在證實臨床療效及安全性之後，藥物才能進行查驗登記並取證上市。整體新藥研發平均約費時10到15年，且同時須擔負龐大之研發經費與承受高失敗率的風險。由於新藥的研發相當耗時，若能縮短早期藥物評估的時間與增進潛力候選藥物篩選的效率，則能加速整體藥物開發流程也可以減少研發成本。 化學相似性定律描述結構相似的化學分子享有相似的化學活性。分子相似性討論的是兩個化學分子在結構上的相似程度，相似性定律是藥物發展的重要理論基礎之一。現今有數個不同的分子相似性模式被用來預測藥物的活性評估，包括該化學分子的藥物動力及代謝分佈等特性。 應用分子相似性的比較以輔助候選藥物活性篩選可提高臨床前藥物開發之成功率，然而目前結合AI系統與分子相似性的比較來有效篩選出候選藥物的分析研究並未被深入探討。本論文研究即是藉找出最佳化學分子相似性特徵之衡量指標，結合人工智慧並將其應用在候選藥物的篩選，輔助藥物開發。期望在新藥探索階段即結合AI以更有效率地預測並篩選出有療效潛力的候選藥物分子結構，不僅可縮短藥物在新藥初期階段的研發時間，並可加速研發流程與減少整體研發成本。

The whole process of new drug development includes the exploration and efficacy study of early drug candidates, as well as preclinical animal tests and clinical human trials for product development. After confirming the clinical efficacy and safety, the drug can then be registered for marketing. It takes an average of 10 to 15 years to develop a new drug, and at the same time has to bear the huge R&D expenditure and the risk of high failure rate. Since the development of new drugs is time consuming, shortening the time for early drug evaluation and drug screening for candidates can not only speed up the process of drug development but also reduce the total cost of research and development. The law of similarity describes that chemical molecules of similar structures share similar chemical activities. Molecular similarity refers to the degree of structural similarity between two chemical molecules. The law of similarity is one of the important theoretical foundations for drug development. Several models of molecular similarity are used during drug development to predict candidate drug activities, including pharmacokinetics, toxicology and metabolic distribution. The application of molecular similarity comparison to assist drug candidate screening can improve the success rate of preclinical drug development. However, the involvement of AI system to effectively improve this process of candidate selection has not been discussed in depth. The research of this thesis is to find a better model of chemical molecular similarity combined with artificial intelligence to assist selection of candidate drugs. It is expected that the combination of AI in the new drug exploration stage can efficiently predict and screen out drug candidates with ideal therapeutic potential, which will not only shorten the time of early stage drug development, but also accelerate the research and development process and reduce the overall research and development costs.