美國北卡羅萊納大學讓AI自行設計化合物，還可指定化合物特性加速藥物開發

美國北卡羅萊納大學讓AI自行設計化合物，還可指定化合物特性加速藥物開發

News from: iThome & UMC

現在藥物開發只能在已知的化合物資料庫中進行虛擬篩選，ReLeaSE為科學家帶來的能力，就像是提供一家雜貨店和一位可以製作任何餐點的私人廚師，擁有創造與評估新化合物的獨特能力。

北卡羅萊納大學將人工智慧應用在藥物開發上，他們開發了名為ReLeaSE的人工智慧系統，能自己重頭開始設計新的藥物分子，該系統有兩個神經網路，分別是老師以及學生代理人，由老師教授學生化學分子的知識後，由學生進行創造新的分子。ReLeaSE現在已經能自動設計具特殊性質的化合物。



論文作者之一Olexandr Isayev教授表示，虛擬篩選是製藥業一項廣泛用於識別候選藥物的方法，進行虛擬篩選的科學家，就像是在餐廳點餐的顧客，可能會受到菜單的限制。而ReLeaSE是虛擬篩選一項強而有力的創新，能為科學家提供一家雜貨店和一位可以製作任何餐點的私人廚師。

而這個雜貨店與廚師的角色，便是ReLeaSE系統中的兩個神經網路老師與學生代理人。老師代理人知道約170萬種已知生物活性分子的化學結構，以及其化學結構字詞的語法與命名規則。而學生向老師學習了分子字母表和語言規則後，學生便有能力開始創造新的字詞或是化學分子，與老師相互合作後，則學生在提出可用於新藥的新分子的表現越來越好。研究團隊表示，ReLeaSE代表著增強學習也能用在結構演進上，而這樣的進展將會大幅加速新藥開發，Olexandr Isayev提到，製藥業常用的虛擬篩選計算方法，可以允許科學家評估現存的大型化學資料庫，但該方法的限制就在於僅能對已知的化學分子進行，而ReLeaSE則具有創造與評估新分子的獨特能力。

研究人員可以設定ReLeaSE生成具有特定性質的化學分子，像是生物活性或是安全性的限制，論文主要作者Mariya Popova強調，他們利用ReLeaSE來設計具有許多不同特性的分子化學資料庫，可以精準控制化合物的物理性質例如熔點或是在水中的溶解度，在發現藥物的研究上，ReLeaSE已經設計出了一種能控制白血病的新化合物。

研究團隊提到，ReLeaSE能設計出特定生物活性與最佳安全性的化學分子，而這個化學分子能馬上申請專利，大幅縮短開發新藥的時程，這對於製藥產業有極大的吸引力。

----------------------------------------------------------------------

An artificial-intelligence approach created at the University of North Carolina at Chapel Hill can teach itself to design new drug molecules from scratch.

ReLeaSE is an algorithm and computer program developed at the UNC Eshelman School of Pharmacy. It comprises two neural networks: a teacher and a student. The teacher knows the syntax and linguistic rules behind the vocabulary of chemical structures for about 1.7 million known biologically active molecules, said K. H. Lee Distinguished Professor Alexander Tropsha, Ph.D., one of the creators of the new AI system.

After learning the molecular alphabet and the rules of the language, the student starts creating new ‘words’, or molecules,” Tropsha said. “If the new word-molecule is realistic and has the desired meaning, the teacher approves. If not, the teacher disapproves, forcing the student to avoid bad words and create good ones.”

By working with the teacher, the student gets better and better at proposing molecules that are likely to be useful as new medicines.

ReLeaSE stands for Reinforcement Learning for Structural Evolution, and its creators say they believe this application of machine learning can dramatically accelerate the design of new drug candidates. The university has applied for a patent for the technology, and the team published a proof-of-concept study in Science Advances.

Tropsha and Research Assistant Professor Olexandr Isayev, Ph.D., are the corresponding authors on the study. The lead author is Mariya Popova, a graduate student the School’s Division of Chemical Biology and Medicinal Chemistry and Laboratory for Molecular Modeling.

Isayev said that ReLeaSE is a powerful innovation to virtual screening, the computational method widely used by the pharmaceutical industry to identify viable drug candidates. Virtual screening allows scientists to evaluate existing large chemical libraries, but the method only works for known chemicals. ReLeASE has the unique ability to create and evaluate new molecules.

“A scientist using virtual screening is like a customer ordering in a restaurant; what can be had is usually limited by the menu,” Isayev said. “We want to give scientists a grocery store and a personal chef who can create any dish they want.”

The UNC-Chapel Hill researchers have been able to use ReLeaSE to generate molecules with properties that they specified, such as desired bioactivity and safety profiles.

“We have used the ReLeaSE method to design chemical libraries of molecules with many different properties,” Popova said. “We were able to customize physical properties, such as melting point and solubility in water, and more relevant to drug discovery, design new compounds with inhibitory activity against Janus protein kinase 2, an enzyme that is associated with leukemia.”

Tropsha said he believes that ReLeaSE should be of great interest to the pharmaceutical industry.

“The ability of the algorithm to design new, and therefore immediately patentable, chemical entities with specific biological activities and optimal safety profile should be highly attractive to an industry that is constantly searching for new approaches to shorten the time it takes to bring a new drug candidate to clinical trials,” he said.

DOI: 10.1126/sciadv.aap7885