憂鬱症是一種可能危及生命之心理衛生疾患。根據世界衛生組織 (WHO,
2012) 推估，到了 2020 年時，其將成為生理失能的第二大主因， 2030 年時則將成為當代工作效益低落的主要影響來源。儘管市面上已存在多樣化的醫療選項，此疾病底層之主要致病機轉仍舊不甚明朗。市售抗憂鬱劑的最主要之兩個問題層面包括療效的延遲以及不如預期，再加上其所伴隨的廣泛副作用，對於未來藥劑之改良毫無疑問地還有很大的進步空間。我們研究的目的在於開發系統模型去預測抗憂鬱劑的淺在副作用和標靶，希望可以藉此對未來的藥物開發和療程提供助益。

我們提出了一個整合式的系統框架，從線上資料庫當中汲取 816 顆藥物和其相關聯之 653 項化學結構、984 個生物特性和 6,111 組副作用檔案來預測未知的副作用和淺在的藥物標靶。從使用的四組機器學習演算法當中，我們發現整合式隨機森林模型所達到的預測結果最為理想，因而更進一步地透過此組模型來做抗憂鬱劑個案的預測研究。研究當中的 15 顆憂鬱關連藥物包含 9 顆第一代、5 顆第二代抗憂鬱劑和一個有著和三環類抗憂鬱劑相似化學結構之肌肉鬆弛劑。對於抗憂鬱劑之副作用以及標靶的預測結果分別得到：AUROC: 0.9140834, AUPR: 0.5185952; AUROC: 0.9513566, AUPR: 0.3101223，在後續文獻核對當中更再次證實了我們預測模型之有效性。

Depression is a life-threatening mental health disorder which is expected to be the second leading cause of psychosocial disability throughout the world by 2020 and will become the largest contributor to lost work productivity by 2030 as reported by World Health Organization (WHO, 2012). Despite the availability of various therapeutic options, the underlying pathological mechanisms remain unclear. The important concerns with
antidepressants are delayed therapeutic response and insufficient efficacy. With a wide range of adverse effects, there is no doubt a large unmet need for better pharmaceutical treatment. The purpose of our study is to develop a computational approach to investigate potential side effects and targets of antidepressants, hoping to provide support for better strategies for the future of drug development and therapy.

We presented an aggregation framework to predict unknown side effects and hidden targets from 816 drugs by adopting 653 chemical, 984 biological and 6,111 phenotypic features. Among four machine learning-based algorithms, we found that the aggregation random forest model achieved best in overall performance. Hence, we used this computational approach to predict the potential candidates for antidepressants. We conducted the case
study using 15 depression-related drugs, including 9 first generation, 5 second generation antidepressants and 1 muscle relaxant that has a structure similar to tricyclic antidepressant (TCA). The in silico model obtained promising results with AUROC score of 0.9140834,
AUPR score of 0.5185952 for side effects prediction and AUROC score of 0.9513566, AUPR score of 0.3101223 for targets prediction.

List of figures v
List of tables vii
Nomenclature viii
1 Introduction 1
1.1 Research background . . . . . . . . . . . . . . . . 1
1.1.1 Drug regulation and development . . . . . . . . . 1
1.1.2 How drugs affect the body (PK/PD studies) . . . . 6
1.1.3 Side effects: The good and the bad . . . . . . . 9
1.1.4 Computational approaches . . . . . . . . . . . . 11
1.2 Related work . . . . . . . . . . . . . . . . . . . 12
1.3 Motivation . . . . . . . . . . . . . . . . . . . . 13
2 Methodology 17
2.1 Experimental design . . . . . . . . . . . . . . . 17
2.2 Materials . . . . . . . . . . . . . . . . . . . . 20
2.2.1 PubChem . . . . . . . . . . . . . . . . . . . . 20
2.2.2 DrugBank . . . . . . . . . . . . . . . . . . . . 21
2.2.3 SIDER . . . . . . . . . . . . . . . . . . . . . 22
2.3 Methods . . . . . . . . . . . . . . . . . . . . . 22
2.3.1 Random forest . . . . . . . . . . . . . . . . . 22
2.3.2 k-Nearest neighbors . . . . . . . . . . . . . . 22
2.3.3 Support vector machines . . . . . . . . . . . . 23
2.3.4 Sparse canonical correlation analysis . . . . . 23
3 Results and performance evaluation 26
3.1 Side effects prediction . . . . . . . . . . . . . 27
3.1.1 General prediction . . . . . . . . . . . . . . . 27
3.1.2 Comparison with Mizutani’s method . . . . . . . 29
3.1.3 Case study: Antidepressants . . . . . . . . . . 31
3.2 Targets prediction . . . . . . . . . . . . . . . . 35
3.2.1 General prediction . . . . . . . . . . . . . . . 35
3.2.2 Case study: Antidepressants . . . . . . . . . . 37
4 Discussion and conclusion 39
References 42
Appendix A The predicted side effects of antidepressants 47
Appendix B The predicted targets of antidepressants 65

[1] drug. (n.d.). Dorland’s Medical Dictionary for Health Consumers., 2007. Retrieved
from http://medical-dictionary.thefreedictionary.com/drug.
[2] drug. (n.d.). Medical Dictionary for the Health Professions and Nursing., 2012. Retrieved
from http://medical-dictionary.thefreedictionary.com/drug.
[3] drug. (n.d.). Mosby’s Medical Dictionary., 2009. Retrieved from http://medicaldictionary.
thefreedictionary.com/drug.
[4] F. Bosch and L. Rosich. The contributions of paul ehrlich to pharmacology: A tribute
on the occasion of the centenary of his nobel prize. Pharmacology, 82(3):171–179,
2008.
[5] J. Lazarou, BH. Pomeranz, and PN. Corey. Incidence of adverse drug reactions in
hospitalized patients: A meta-analysis of prospective studies. JAMA, 279(15):1200–
1205, 1998.
[6] K. Jarell. Regulatory history: Elixir sulfanilamide. Journal of GXP Compliance,
16(3):12, 2012.
[7] F. O. Kelsey. Problems raised for the fda by the occurrence of thalidomide embryopathy
in germany, 1960-1961. American Journal of Public Health and the Nations Health,
55(5):703–707, 1965.
[8] J. H. Kim and A. R. Scialli. Thalidomide: The tragedy of birth defects and the effective
treatment of disease. Toxicological Sciences, 122(1):1–6, 2011.
[9] A. Mullard. New drugs cost us [dollar] 2.6 billion to develop. Nature Reviews Drug
Discovery, 13(12):877–877, 2014.
[10] The biopharm guide to biopharmaceutical development. supplement to biopharm. 2nd
ed. 2002.
[11] P. Barber and D. Robertson. Essentials of pharmacology for nurses. 2012.
[12] A. E. Routes. Drug absorption, distribution and elimination; pharmacokinetics.
http://www.columbia.edu/itc/gsas/g9600/2004/GrazianoReadings/Drugabs.pdf.
[13] S. M. Strittmatter. Old drugs learn new tricks. Nature Medicine, 20(6):590–591, 2014.
[14] M. S. Boguski, K. D. Mandl, and V. P. Sukhatme. Repurposing with a difference.
Science, 324(5933):1394–1395, 2009.
[15] I. R. Edwards and J. K. Aronson. Adverse drug reactions: definitions, diagnosis, and
management. The Lancet, 356(9237):1255–1259, 2000.
[16] K. M. Giacomini, R. M. Krauss, D. M. Roden, M. Eichelbaum, M. R. Hayden, and
Y. Nakamura. When good drugs go bad. Nature, 446(7139):975–977, 2007.
[17] M. J. Alomar. Factors affecting the development of adverse drug reactions (Review
article). Saudi Pharmaceutical Journal: SPJ, 22(2):83–94, 2014.
[18] X. Chen, C. C. Yan, X. Zhang, X. Zhang, F. Dai, J. Yin, and Y. Zhang. Drug–target
interaction prediction: databases, web servers and computational models. Briefings in
bioinformatics, bbv066, 2015.
[19] A. Bender, J. Scheiber, M. Glick, J. W. Davies, K. Azzaoui, J. Hamon, ..., and J. L.
Jenkins. Analysis of pharmacology data and the prediction of adverse drug reactions
and off-target effects from chemical structure. ChemMedChem, 2(6):861–873, 2007.
[20] M. Campillos, M. Kuhn, A. C. Gavin, L. J. Jensen, and P. Bork. Drug target identification
using side-effect similarity. Science, 321(5886):263–266, 2008.
[21] L. Xie, J. Li, L. Xie, and P. E. Bourne. Drug discovery using chemical systems biology:
identification of the protein-ligand binding network to explain the side effects of cetp
inhibitors. PLoS Comput Biol, 5(5), 2009. e1000387.
[22] N. P. Tatonetti, G. H. Fernald, and R. B. Altman. A novel signal detection algorithm
for identifying hidden drug-drug interactions in adverse event reports. Journal of the
American Medical Informatics Association, 19(1):79–85, 2012.
[23] S. Mizutani, E. Pauwels, V. Stoven, S. Goto, and Y. Yamanishi. Relating drug–protein
interaction network with drug side effects. Bioinformatics, 28(18):i522–i528, 2012.
[24] M. Takarabe, M. Kotera, Y. Nishimura, S. Goto, and Y. Yamanishi. Drug target prediction
using adverse event report systems: a pharmacogenomic approach. Bioinformatics,
28(18):i611–i618, 2012.
[25] M. Liu, Y. Wu, Y. Chen, J. Sun, Z. Zhao, X. W. Chen, ..., and H. Xu. Large-scale prediction
of adverse drug reactions using chemical, biological, and phenotypic properties
of drugs. Journal of the American Medical Informatics Association, 19(e1):e28–e35,
2012.
[26] B. Bondy. Pathophysiology of depression and mechanisms of treatment. Dialogues in
clinical neuroscience, 4:7–20, 2002.
[27] T. C. Baghai, C. Zirngibl, B. Heckel, N. Sarubin, and R. Rupprecht. Individualized pharmacological
treatment of depressive disorders state of the art and recent developments.
Journal of Depression and Anxiety, 2014.
[28] R. C. Team. R: A language and environment for statistical computing. 2013.
[29] Pauwels, Edouard, Stoven, Véronique, and Y. Yamanishi. Predicting drug side-effect
profiles: a chemical fragment-based approach. BMC Bioinformatics, 12(1):1–13, 2011.
[30] S. Kim, P. A. Thiessen, E. E. Bolton, J. Chen, G. Fu, A. Gindulyte, ..., and J. Wang.
PubChem substance and compound databases. Nucleic acids research, 2015. gkv951.
[31] V. Law, C. Knox, Y. Djoumbou, T. Jewison, A. C. Guo, Y. Liu, ..., and A. Tang.
DrugBank 4.0: shedding new light on drug metabolism. Nucleic acids research, 42(D1),
2014. D1091-D1097.
[32] M. Kuhn, I. Letunic, L. J. Jensen, and P. Bork. The SIDER database of drugs and side
effects. Nucleic acids research, 2015. gkv1075.
[33] Y. Cao, A. Charisi, L. C. Cheng, T. Jiang, and T. Girke. ChemmineR: a compound
mining framework for R. Bioinformatics, 24(15):1733–1734, 2008.
[34] UniProt Consortium. UniProt: a hub for protein information. Nucleic acids research,
2014. gku989.
[35] L. Breiman. randomForest: Breiman and Cutler’s random forests for classification and
regression. 2006.
[36] A. Karatzoglou, A. Smola, K. Hornik, and A. Zeileis. kernlab-an S4 package for kernel
methods in R. 2004.
[37] H. Hotelling. Relations between two sets of variates. Biometrika, 28(3/4):321–377,
1936.
[38] I. González, S. Déjean, P. G. Martin, and A. Baccini. CCA: An R package to extend
canonical correlation analysis. Journal of Statistical Software, 23(12):1–14, 2008.
[39] D. Witten, R. Tibshirani, S. Gross, B. Narasimhan, and M. D. Witten. Package ‘pma’.
Genetics and Molecular Biology, 8(1):28, 2013.
[40] T. Sing, O. Sander, N. Beerenwinkel, and T. Lengauer. ROCR: visualizing classifier
performance in R. Bioinformatics, 21(20):3940–3941, 2005.
[41] M. Kuhn. Caret package. Journal of Statistical Software, 28(5), 2008.
[42] W. Zhang, F. Liu, L. Luo, and J. Zhang. Predicting drug side effects by multi-label
learning and ensemble learning. BMC bioinformatics, 16(1):1, 2015.
[43] H. Chen, J. Zha, L. Yuan, and Z. Wang. Effects of fluoxetine on behavior, antioxidant
enzyme systems, and multixenobiotic resistance in the Asian clam Corbicula fluminea.
Chemosphere, 119:856–862, 2015.
[44] H. Möhler. The GABA system in anxiety and depression and its therapeutic potential.
Neuropharmacology, 62(1):42–53, 2012.
[45] R. Cacabelos. Pharmacogenomics of central nervous system (CNS) drugs. Drug
Development Research, 73(8):461–476, 2012.
[46] M. M. Gutierrez, J. Rosenberg, and W. Abramowitz. An evaluation of the potential
for pharmacokinetic interaction between escitalopram and the cytochrome P450 3A4
inhibitor ritonavir. Clinical therapeutics, 25(4):1200–1210, 2003.
[47] M. Szewczuk-Bogusławska, A. Kiejna, M. Grzesiak, J. A. Beszłej, I. Chlebowska,
K. Orzechowska-Juzwenko, and P. Milejski. The influence of clomipramine on CYP2D6
activity. Psychiatria polska, 41(2):243–249, 2006.
[48] K. K. Nielsen, J. P. Flinois, P. Beaune, and K. Brøsen. The biotransformation of
clomipramine in vitro, identification of the cytochrome P450s responsible for the
separate metabolic pathways. Journal of Pharmacology and Experimental Therapeutics,
277(3):1659–1664, 1996.
[49] M. Okubo, N. Murayama, J. Miura, Y. Chiba, and H. Yamazaki. Effects of cytochrome
P450 2D6 and 3A5 genotypes and possible coadministered medicines on the metabolic
clearance of antidepressant mirtazapine in Japanese patients. Biochemical pharmacology,
93(1):104–109, 2015.
[50] E. Lounkine, M. J. Keiser, S. Whitebread, D. Mikhailov, J. Hamon, J. L. Jenkins, ...,
and B. K. Shoichet. Large-scale prediction and testing of drug activity on side-effect
targets. Nature, 486(7403):361–367, 2012.
[51] C. J. O’Donnell, K. Grime, P. Courtney, D. Slee, and R. J. Riley. The development
of a cocktail CYP2B6, CYP2C8, and CYP3A5 inhibition assay and a preliminary
assessment of utility in a drug discovery setting. Drug metabolism and disposition,
35(3):381–385, 2007.
[52] Y. Kitamura, Y. Fujitani, K. Kitagawa, T. Miyazaki, H. Sagara, H. Kawasaki, ..., and
Y. Gomita. Effects of imipramine and bupropion on the duration of immobility of
ACTH-treated rats in the forced swim test: involvement of the expression of 5-HT2A
receptor mRNA. Biological and Pharmaceutical Bulletin, 31(2):246–249, 2008.
[53] A. Martin, L. Scahill, and C. Kratochvil. Pediatric Psychopharmacology: Principles
and Practice. Oxford University Press, 2011.
[54] M. Kotlyar, L. H. Brauer, T. S. Tracy, D. K. Hatsukami, J. Harris, C. A. Bronars,
and D. E. Adson. Inhibition of CYP2D6 activity by bupropion. Journal of clinical
psychopharmacology, 25(3):226–229, 2005.
[55] D. J. Stein, B. Lerer, and S. M. Stahl. Evidence-based psychopharmacology. New York:
Cambridge University Press., 2005.
[56] M. El Mansari, R. Ghanbari, S. Janssen, and P. Blier. Sustained administration of
bupropion alters the neuronal activity of serotonin, norepinephrine but not dopamine
neurons in the rat brain. Neuropharmacology, 55(7):1191–1198, 2008.
[57] E. C. Lauterbach. Psychotropic drug effects on gene transcriptomics relevant to parkinson’s
disease. Progress in Neuro-Psychopharmacology and Biological Psychiatry,
38(2):107–115, 2012.
[58] V. C. Bortoli, R. L. Nogueira, and H. Zangrossi Jr. Effects of fluoxetine and buspirone
on the panicolytic-like response induced by the activation of 5-HT1A and 5-HT2A
receptors in the rat dorsal periaqueductal gray. Psychopharmacology, 183(4):422–428,
2006.
[59] C. Spindelegger, R. Lanzenberger, W. Wadsak, L. K. Mien, P. Stein, M. Mitterhauser,
..., and S. Kasper. Influence of escitalopram treatment on 5-HT1A receptor binding in
limbic regions in patients with anxiety disorders. Molecular psychiatry, 14(11):1040–
1050, 2009.