基於生物醫學文獻建構藥物新適應症預測模型

Literature-based Discovery for Drug Repositioning: A Predicate-Pattern-Importance Ranking Approach

陳昀靖(Yun-Jing Chen)

指導教授 ： 魏志平

國立臺灣大學/管理學院/資訊管理學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201903652

摘要

開發新藥物成本極高且風險極大，因此研究人員開始尋找許多替代方案。新藥上市有了新選擇——「舊藥新用」：從已經被核准做為臨床使用的藥物中，發現新的適應症用途。因舊藥已有許多完整的臨床藥物資訊及人體使用安全性資料，因此比起研發新藥，不但能夠縮短開發時間也能降低研發成本及風險。
1986年，Swanson利用資訊檢索與資料探勘技術，透過醫療文獻中的關係建構出醫療關係網路，希望能透過此網路實現舊藥新用，之後許多學者沿用此方法。我們沿用前一篇方法 (Lee, 2017) 提出的路徑重要性分類模型，從含有語意的醫療關係所建構出的網路中，利用機器學習技術透過已知的醫療語意關係，將這些醫療元件及關係轉換成富含資訊的向量，藉此擷取出含有醫療含義的特徵值，判斷此路徑對於某焦點藥物與候選疾病是否重要，針對某焦點藥物，依照分類的結果計算分數，對其候選疾病進行排序。
實驗證明我們加入的醫療語意向量能顯著提升原本的路徑重要性分類模型，藉此更有效找出潛在藥物新適應症。

關鍵字

舊藥新用 ； 語義關係 ； 監督式學習 ； 路徑重要性分類模型 ； 知識圖譜

並列摘要

Drug development is extremely costly and risky, so researchers are looking for alternative approaches. There’s a new approach: Drug Repurposing, using existing drugs approved by FDA to find new indications. Compared with the development of new drugs, existing drugs have more complete clinical drug information and human safety data. It not only can shorten development time but also reduce the risks on R&D.
In 1986, Swanson proposed an approach using information retrieval and text mining techniques to construct a biomedical network composed of links between biomedical concepts from biomedical literatures. Through analyzing this network, drug repurposing can be achieved. Many researchers have followed this approach. Previous studies (Lee, 2017) proposed a path-importance-based approach. We follow this approach. In our research, we use machine learning techniques to convert biomedical entities and relations into representative biomedical vectors, discriminate whether a path is important or not and decide the candidate diseases given a focal drug.
The empirical results show that the representative biomedical vectors can significantly improve the path-importance-based classification, which in turn can support effective drug repurposing.

並列關鍵字

drug repurposing ； semantic predication ； supervised learning ； path importance classification ； knowledge graph

參考文獻

Ashburn, T. T., & Thor, K. B. (2004). Drug repositioning: Identifying and developing new uses for existing drugs. Nature Reviews Drug Discovery, 3(8), 673–683.

Boolell, M., Allen, M. J., Ballard, S. A., Gepi-Attee, S., Muirhead, G. J., Naylor, A. M., … Gingell, C. (1996). Sildenafil: an orally active type 5 cyclic GMP-specific phosphodiesterase inhibitor for the treatment of penile erectile dysfunction. International Journal of Impotence Research, 8(2), 47—52.

Bordes, A., Usunier, N., Weston, J., & Yakhnenko, O. (2013). TransE: Translating Embeddings for Modeling Multi-Relational Data. Advances in NIPS, 26, 2787–2795.

Bordes, A., & Weston, J. (2011). Learning Structured Embeddings of Knowledge Bases. (Twenty-Fifth AAAI Conference on Artificial Intelligence.), 301–306.

Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. In Proceedings of the 22nd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining, ACM, 785–794.