Integrative Research of Neuropharmacology and Informatics Pharmacology for Mental Disorder

Abstract

Mental illness poses a huge social burden, accounting for approximately 14% of all deaths. Depression, a major component of mental illness, affects approximately 300 million people worldwide, mainly in developed countries, and is not only a major social burden but also a cause of suicide. The social burden of depression is estimated to increase further in developing countries, and overcoming it is a pressing issue for all countries, including Japan. Although clinical evidence has demonstrated the efficacy of serotonergic neurotransmission enhancers in the treatment of depression, the full picture of their therapeutic effects has not yet been fully elucidated. In this review, we show that the hyperactivity of serotonin neurons, especially those in the dorsal raphe nucleus, is commonly induced by various antidepressants within a period corresponding to the onset of their clinical efficacy. We established quantitative prediction methods for pharmacological activity using only chemical structures to translate the biological understanding of mental disorders, including major depressive disorders, into clinically effective therapeutics. Our method exhibited better performance than the previously reported methods of quantitative prediction, while targeting a larger number of proteins. Our article suggests the importance of integrative neuropharmacology and informatics-based pharmacology studies to understand the biological basis of mental disorders and facilitate drug development for these disorders.

1. INTRODUCTION

Mental illness poses a huge social burden, accounting for approximately 14% of all deaths.¹⁾ Depression is a major component of mental illness that affects approximately 300 million people worldwide,²⁾ mainly in developed countries, and is a major social burden and cause of suicide. As developing countries become more advanced, the social burden of depression will further increase; overcoming it is a pressing issue for all countries, including Japan. The antidepressant effect of imipramine was discovered in the 1950s,³⁾ which led to the development of many antidepressants, including selective serotonin reuptake inhibitors (SSRIs), serotonin, and noradrenaline reuptake inhibitors (SNRIs).^4,5) However, their therapeutic effects remain unclear. Imipramine enhances serotonin neurotransmission by inhibiting serotonin reuptake. Most antidepressants potentiate serotonergic neurotransmission.^6–8) Serotonin depletion causes symptom recurrence in patients with depression.⁹⁾ The hypothesis that enhancing serotonin neurotransmission leads to the treatment of depression has been the basis for antidepressant drug development for many years; this hypothesis has been supported by the many studies but does not explain the delayed onset of drug effects and treatment-resistance in some patients.¹⁰⁾ Therefore, better biological understanding of the recovery process from antidepressant drugs is essential to develop new drugs to overcome these problems.

2. ENHANCEMENT OF SEROTONERGIC NEUROTRANSMISSION VIA HYPERACTIVITY OF SEROTONIN NEURONS

Two possible mechanisms for enhancing serotonin neurotransmission are 1) inhibition of the serotonin transporter (SERT), which is responsible for the reuptake of released serotonin, and 2) inhibition of the serotonin-degrading enzyme monoamine oxidases. Inhibitors of SERT and monoamine oxidases increase extracellular serotonin concentrations^11–13) and are widely used as antidepressants.^7,8) The potentiating effect of serotonin neurotransmission by serotonin reuptake and degradation enzyme inhibition, the points of action of conventional drugs, theoretically occurs by prolonging the time during which the released serotonin is available for action. Thus, the potentiating effects are highly dependent on the activity of the serotonergic neurons responsible for serotonin release. Activity of serotonin neurons in the dorsal raphe nucleus (DRN) is reduced in animal models of depression, such as chronic social defeat stress,^14,15) suggesting that the effects of conventional drugs are reduced in stressed animals and possibly in patients with major depressive disorder (MDD) owing to decreased serotonergic activity (Fig. 1).

Fig. 1. Effects of Conventional Antidepressants Depend on the Neuronal Activity of Serotonin Neurons

SERT: serotonin transporter, SSRI: selective serotonin reuptake inhibitor.

To investigate whether serotonergic activity is the real target of conventional antidepressants, we previously analyzed the effects of chronic treatment with conventional antidepressants on serotonergic activity using cultured brain slices containing serotonergic neurons.^16,17) We found that chronic treatment with conventional antidepressants, such as citalopram (an SSRI), increased the extracellular serotonin concentration in a time-dependent manner, which was inhibited by the pharmacological inhibition of voltage-gated sodium channels and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type glutamate receptors as well as the removal of extracellular calcium. These data strongly indicate that the increased extracellular serotonin concentration is due to increased activity of serotonergic neurons.¹⁶⁾ Interestingly, antidepressants, such as SNRIs and tricyclics, commonly induce an increase in the activity of serotonergic neurons.¹⁷⁾ To examine whether this increase is induced in vivo, we analyzed the activity of serotoninergic neurons in the DRN after chronic SSRI treatment. We found that the spontaneous activity of the DRN serotonergic neurons after chronic SSRI treatment was significantly higher than that after vehicle treatment.¹⁸⁾ Collectively, these data suggest that chronic treatment with conventional antidepressants increases serotonin neuronal activity.

We then explored whether and how antidepressants are effective, at least in part, for treatment-resistant depression and how they affect the activity of serotonin neurons. The clinical literature indicates that some antipsychotics, such as olanzapine, augment the efficacy of conventional antidepressants. A combination of olanzapine and fluoxetine is used to treat treatment-resistant depression.¹⁹⁾ Using cultured brain slices, we found that acute treatment with olanzapine significantly augmented the effects of SSRI on extracellular serotonin levels.²⁰⁾ Notably, this augmentation was occluded in the presence of a γ-aminobutyric acid (GABA)_A or 5-hydroxytryptamine type 6 (5-HT₆) receptor antagonist, indicating the possible involvement of disinhibition through 5-HT₆ receptor antagonism. Further electrophysiological experiments revealed that olanzapine and 5-HT₆ receptor antagonists commonly inhibit GABAergic neurons in the DRN, which in turn leads to the activation of DRN serotonin neurons.²⁰⁾

Ketamine, an N-methyl-D-aspartate (NMDA)-type glutamate receptor antagonist, has been used as an anesthetic agent; however, recent clinical studies have demonstrated that ketamine has robust and rapid antidepressant effects.^21,22) Using microdialysis, we found that ketamine rapidly increased extracellular serotonin levels in the medial prefrontal cortex through the activation of AMPA-type glutamate receptors in the DRN.²³⁾ Further analysis using local perfusion of ketamine revealed that ketamine activates cholinergic projections from the pedunculopontine tegmental nucleus to the DRN, which indirectly activates glutamatergic inputs to DRN serotonin neurons, leading to a rapid increase in extracellular serotonin.²⁴⁾ Collectively, our data suggest that compounds effective for treatment-resistant depression and conventional antidepressants commonly induce hyperactivity of serotonergic neurons, especially in the DRN.

3. OPTOGENETIC MANIPULATION OF DRN SEROTONIN NEURONS

Next, we tested whether the artificially induced hyperactivity of serotonin neurons in the DRN alone produced antidepressant effects. However, conventional stimulation by electrodes or lesion experiments are not capable of cell type-specific manipulation. Optogenetics, in which neural activity is controlled by light, is an innovative neuroscientific technology.²⁵⁾ It utilizes light-sensitive ion channels and pumps, such as channel rhodopsin 2 (ChR2), to manipulate neuronal activity in a cell type-specific manner; however, its application to serotonergic neurons requires technology to selectively express large amounts of transgenes in serotonin neurons. To accomplish this, we isolated a new serotonin neuron-selective promoter, combined it with a unique transcriptional amplification mechanism, and succeeded in developing a virus with both extremely high selectivity (95%) and high expression, sufficient to control the activity of serotonin neurons (Figs. 2A, B). We then generated mice and rats in which serotonergic activity could be enhanced by light using the developed virus and found that enhancement of serotonergic activity leads to rapid antidepressant-like effects.²⁶⁾ Furthermore, serotonergic hyperactivity rapidly induces hedonic responses that are often impaired in depressed patients.²⁷⁾ These findings strongly support our hypothesis that depression treatment targets the activity of the serotonergic nerve.

Fig. 2. Viral Vectors Capable of Highly Specific and Potent Gene Expression in Serotonin Neurons

(A) Immunohistochemical analysis of green fluorescent protein (GFP) and tryptophan hydroxylase (TPH), a marker for serotonin neurons, in the dorsal raphe nucleus (DRN) of mice injected with serotonin neuron-selective lentiviral vectors bearing Venus (smTPH2::Venus). Scale bar = 20 µm. (B) Cells with enhanced yellow fluorescent protein (eYFP) tagged to ChETA (excitatory optogenetic actuator) and eArchT (inhibitory optogenetic actuator) were analyzed using a whole-cell patch clamp. Blue light (20 pulses, 20 Hz, 5 ms) or green light (1 pulse, 0.5 or 1 s) was applied through fiber optics to illuminate the patched cells. Current-clamp recording (upper panel) and voltage-clamp recording (lower panel) are shown. Reproduced from Ref. 26.

4. QUANTITATIVE PREDICTION OF PHARMACOLOGICAL ACTIVITY USING GRAPH CONVOLUTIONAL NEURAL NETWORKS (GCNS)

Previous studies have revealed heterogeneity in serotoninergic neurons in the DRN using single-cell RNA-seq technology.^28–30) Moreover, in vivo endoscopic analysis has demonstrated the functional diversity of DRN serotonin neurons; for instance, some DRN serotonin neurons respond to reward while others part of them are activated by aversive stimuli.³¹⁾ These results indicate that the pharmacological targeting of receptors or channels specifically expressed in a subpopulation of DRN serotonin neurons is a promising way to develop antidepressants. However, it is impractical to screen for several pharmacological targets identified by biological analyses, including single-cell RNA-seq. In recent years, remarkable progress has been made in information science, represented by deep learning. Convolutional neural networks (CNNs) and their descendants with aid of general-purpose computing on graphics processing units have revolutionized the many areas of research, including computer vision and image processing.^32,33) Conventional CNN can treat graphs with fixed structures between nodes; that is, each node is connected to four or six nodes in two-dimensional and three-dimensional images, respectively. However, conventional CNN cannot handle graphs with variable structures between the nodes, such as social networks or chemical structures. Duvenaud et al. expanded the CNN to graphs with variable structures called GCN.³⁴⁾ More recently, Wu et al. demonstrated the superior performance of GCNs in predicting various molecular characteristics, such as lipophilicity, biological activity, and toxicity.³⁵⁾ Then, we hypothesized that GCNs can learn the relationships between chemical structure and pharmacological activity if a large set of “real” assay data for pharmacological activity of a variety of chemical structures is available. To address this, we focused on ChEMBL,³⁶⁾ a database that accumulates structure–activity relationship studies, and developed a method to quantitatively predict the pharmacological effects of arbitrary compounds by combining them with GCNs. As a result, we succeeded in developing the world’s best quantitative pharmacological action prediction method, which can predict pharmacological actions with higher accuracy than previously reported methods³⁷⁾ and the number of target proteins was five-fold higher (25 vs. 127). To examine the robustness of the constructed prediction models, we performed in silico screening targeting SERT and found that a compound, 5-Chloro-2-piperidin-4-yl-1,3-benzothiazole hydrochloride (CHEMBL1377753) (1) (Fig. 3A), is predicted to potently inhibit SERT (predicted IC₅₀ = 10 nM). We then measured the IC₅₀ value of the compound 1 in HEK293T cells expressing human SERT and found that the compound 1 has IC₅₀ value of 6.24 nM for human SERT. Finally, we investigated whether this compound has antidepressant-like efficacy in vivo and found that acute administration of compound 1 significantly decreased the immobility time in the mouse tail suspension test but did not significantly affect locomotor activity (Figs. 3B, C). Collectively, these results strongly indicate the robustness and usefulness of the constructed prediction models for pharmacological activity.³⁸⁾

Fig. 3. Validation of the Prediction Models for Pharmacological Activity against Serotonin Transporter

(A) Chemical structure of CHEMBL1377753 (1). (B, C) Administration of 1 induced antidepressant-like effects in mice. After intraperitoneal injection of 1 (1, 10 mg/kg), the immobility duration was measured in a tail suspension test (B) and travelled distance was measured in an open field test (C). Data represent the mean ± standard error of the mean (S.E.M). * p < 0.05 vs. saline group. n = 6–8 mice per group. Reproduced from Ref. 38.

5. FUTURE PERSPECTIVES AND CONCLUSIONS

Serotonin neurons are associated with various mental disorders, such as MDD, schizophrenia, and anxiety disorders.⁶⁾ Reagents affecting serotonergic neurotransmission are widely used to treat such disorders.^6–8) Clinical evidence has demonstrated the efficacy of serotonergic neurotransmission enhancers for MDD treatment^7,8); however, their therapeutic effects remain ambiguous. Here, we discussed that the hyperactivity of serotonin neurons, especially those in the DRN, is commonly induced by various antidepressants within a period corresponding to the onset of their clinical efficacy,^{8,16–18,20,24)} indicating the direct activation of DRN serotonin neurons as a potential rapid-onset antidepressant strategy. Single-cell RNA-seq analyses have shown that the DRN serotonin neurons express several receptors and ion channels^28,29); therefore, pharmacological interventions targeting these molecular targets may overcome the delayed onset and treatment resistance associated with current MDD treatments.

To develop compounds specifically acting on the molecular targets, comprehensive pharmacological assays are necessary, which require enormous labor and are often impractical. As described above, we performed integrative research on neuropharmacology and informatics-based pharmacology for mental disorders. Considering the rapid growth in pharmacological, chemical, genomic, and clinical big data,³⁹⁾ timely utilization of information technology is necessary to better understand the mechanisms by which drugs function in human beings. Several studies have successfully identified possible therapeutics by analyzing clinical big data.^40–45) Moreover, we demonstrated the applicability of association rule mining, a rule-based machine learning method, for the early detection of adverse drug reaction signals.⁴⁶⁾ These reports highlight the importance of integrative research in the fields of pharmacology and informatics. However, clinical big data analysis cannot be applied to new compounds because the current epidemiological analysis of clinical big data is based on changes in the frequency of clinical events in the presence or absence of a drug of interest. Pharmacological activity-based prediction is a promising method to expand the applicability of clinical big data analysis to the prediction of newly designed compounds for the detection of clinical events. Construction of prediction methods for comprehensive pharmacological activities and clinical events will aid in predicting the effects of new compounds in humans based on clinical information; however, comprehensive “real” measurement of pharmacological activity for several targets is impractical. Our prediction model, as shown above, is capable of the comprehensive in silico prediction of pharmacological activity with robustness. Therefore, construction of new prediction methods for clinical events based on pharmacological activity and their combined use with our prediction model can facilitate rapid development of drugs with few adverse effects for the treatment of various mental disorders.

Acknowledgments

We wish to thank Professor Emeritus Shuji Kaneko (Kyoto University, Kyoto, Japan) and Professor Hitoshi Hashimoto (Osaka University, Osaka, Japan) for their continued support to our works. We are also grateful for long-lasting collaborations with our laboratory members and outside of our laboratory, especially Prof. Sergey Kasparov (University of Bristol, Bristol, U.K.), Prof. Katsuyuki Kaneda, Dr. Naoya Nishitani (Kanazawa University, Kanazawa, Japan), and Dr. Yu Ohmura (Chinese Institute for Brain Research, Beijing, China).

Conflict of Interest

The author declares no conflict of interest.

Notes

This review of the author’s work was written by the author upon receiving the 2023 Pharmaceutical Society of Japan Award for Young Scientists.

REFERENCES

• 1) Walker ER, McGee RE, Druss BG. Mortality in mental disorders and global disease burden implications a systematic review and meta-analysis. JAMA Psychiatry, 72, 334–341 (2015).
• 2) Vos T, Allen C, Arora M, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study. Lancet, 388, 1545–1602 (2016).
• 3) Brown WA, Rosdolsky M. The clinical discovery of imipramine. Am. J. Psychiatry, 172, 426–429 (2015).
• 4) Hyttel J. Citalopram—pharmacological profile of a specific serotonin uptake inhibitor with antidepressant activity. Prog. Neuropsychopharmacol. Biol. Psychiatry, 6, 277–295 (1982).
• 5) Ansseau M, von Frenckell R, Mertens C, de Wilde J, Botte L, Devoitille JM, Evrard JL, De Nayer A, Darimont P, Dejaiffe G, Mirel J, Meurice E, Parent M, Couzinier JP, Demarez JP, Serre C. Controlled comparison of two doses of milnacipran (F 2207) and amitriptyline in major depressive inpatients. Psychopharmacology (Berl.), 98, 163–168 (1989).
• 6) Lucki I. The spectrum of behaviors influenced by serotonin. Biol. Psychiatry, 44, 151–162 (1998).
• 7) Bonnet U. Moclobemide: therapeutic use and clinical studies. CNS Drug Rev., 9, 97–140 (2003).
• 8) Cipriani A, Furukawa TA, Salanti G, Geddes JR, Higgins JP, Churchill R, Watanabe N, Nakagawa A, Omori IM, McGuire H, Tansella M, Barbui C. Comparative efficacy and acceptability of 12 new-generation antidepressants: a multiple-treatments meta-analysis. Lancet, 373, 746–758 (2009).
• 9) Smith KA, Fairburn CG, Cowen PJ. Relapse of depression after rapid depletion of tryptophan. Lancet, 349, 915–919 (1997).
• 10) Otte C, Gold SM, Penninx BW, Pariante CM, Etkin A, Fava M, Mohr DC, Schatzberg AF. Major depressive disorder. Nat. Rev. Dis. Primers, 2, 16065 (2016).
• 11) Perry KW, Fuller RW. Effect of fluoxetine on serotonin and dopamine concentration in microdialysis fluid from rat striatum. Life Sci., 50, 1683–1690 (1992).
• 12) David DJP, Bourin M, Jego G, Przybylski C, Jolliet P, Gardier AM. Effects of acute treatment with paroxetine, citalopram and venlafaxine in vivo on noradrenaline and serotonin outflow: a microdialysis study in Swiss mice. Br. J. Pharmacol., 140, 1128–1136 (2003).
• 13) Kitaichi Y, Inoue T, Nakagawa S, Boku S, Koyama T. Effects of combined treatment with clorgyline and selegiline on extracellular noradrenaline and serotonin levels. Acta Neuropsychiatr., 24, 369–373 (2012).
• 14) Espallergues J, Teegarden SL, Veerakumar A, Boulden J, Challis C, Jochems J, Chan M, Petersen T, Deneris E, Matthias P, Hahn CG, Lucki I, Beck SG, Berton O. HDAC6 regulates glucocorticoid receptor signaling in serotonin pathways with critical impact on stress resilience. J. Neurosci., 32, 4400–4416 (2012).
• 15) Challis C, Boulden J, Veerakumar A, Espallergues J, Vassoler FM, Pierce RC, Beck SG, Berton O. Raphe GABAergic neurons mediate the acquisition of avoidance after social defeat. J. Neurosci., 33, 13978–13988, 13988a (2013).
• 16) Nagayasu K, Yatani Y, Kitaichi M, Kitagawa Y, Shirakawa H, Nakagawa T, Kaneko S. Utility of organotypic raphe slice cultures to investigate the effects of sustained exposure to selective 5-HT reuptake inhibitors on 5-HT release. Br. J. Pharmacol., 161, 1527–1541 (2010).
• 17) Nagayasu K, Kitaichi M, Nishitani N, Asaoka N, Shirakawa H, Nakagawa T, Kaneko S. Chronic effects of antidepressants on serotonin release in rat raphe slice cultures: high potency of milnacipran in the augmentation of serotonin release. Int. J. Neuropsychopharmacol., 16, 2295–2306 (2013).
• 18) Asaoka N, Nishitani N, Kinoshita H, Kawai H, Shibui N, Nagayasu K, Shirakawa H, Nakagawa T, Kaneko S. Chronic antidepressant potentiates spontaneous activity of dorsal raphe serotonergic neurons by decreasing GABAB receptor-mediated inhibition of L-type calcium channels. Sci. Rep., 7, 13609 (2017).
• 19) Bobo WV, Shelton RC. Olanzapine and fluoxetine combination therapy for treatment-resistant depression: review of efficacy, safety, and study design issues. Neuropsychiatr. Dis. Treat., 5, 369–383 (2009).
• 20) Asaoka N, Nagayasu K, Nishitani N, Yamashiro M, Shirakawa H, Nakagawa T, Kaneko S. Olanzapine augments the effect of selective serotonin reuptake inhibitors by suppressing GABAergic inhibition via antagonism of 5-HT₆ receptors in the dorsal raphe nucleus. Neuropharmacology, 95, 261–268 (2015).
• 21) Berman RM, Cappiello A, Anand A, Oren DA, Heninger GR, Charney DS, Krystal JH. Antidepressant effects of ketamine in depressed patients. Biol. Psychiatry, 47, 351–354 (2000).
• 22) Zarate CA Jr, Singh JB, Carlson PJ, Brutsche NE, Ameli R, Luckenbaugh DA, Charney DS, Manji HK. A randomized trial of an N-methyl-D-aspartate antagonist in treatment-resistant major depression. Arch. Gen. Psychiatry, 63, 856–864 (2006).
• 23) Nishitani N, Nagayasu K, Asaoka N, Yamashiro M, Shirakawa H, Nakagawa T, Kaneko S. Raphe AMPA receptors and nicotinic acetylcholine receptors mediate ketamine-induced serotonin release in the rat prefrontal cortex. Int. J. Neuropsychopharmacol., 17, 1321–1326 (2014).
• 24) Kinoshita H, Nishitani N, Nagai Y, Andoh C, Asaoka N, Kawai H, Shibui N, Nagayasu K, Shirakawa H, Nakagawa T, Kaneko S. Ketamine-induced prefrontal serotonin release is mediated by cholinergic neurons in the pedunculopontine tegmental nucleus. Int. J. Neuropsychopharmacol., 21, 305–310 (2018).
• 25) Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K. Millisecond-timescale, genetically targeted optical control of neural activity. Nat. Neurosci., 8, 1263–1268 (2005).
• 26) Nishitani N, Nagayasu K, Asaoka N, Yamashiro M, Andoh C, Nagai Y, Kinoshita H, Kawai H, Shibui N, Liu B, Hewinson J, Shirakawa H, Nakagawa T, Hashimoto H, Kasparov S, Kaneko S. Manipulation of dorsal raphe serotonergic neurons modulates active coping to inescapable stress and anxiety-related behaviors in mice and rats. Neuropsychopharmacology, 44, 721–732 (2019).
• 27) Nagai Y, Takayama K, Nishitani N, Andoh C, Koda M, Shirakawa H, Nakagawa T, Nagayasu K, Yamanaka A, Kaneko S. The role of dorsal raphe serotonin neurons in the balance between reward and aversion. Int. J. Mol. Sci., 21, 2160 (2020).
• 28) Okaty BW, Freret ME, Rood BD, Brust RD, Hennessy ML, deBairos D, Kim JC, Cook MN, Dymecki SM. Multi-scale molecular deconstruction of the serotonin neuron system. Neuron, 88, 774–791 (2015).
• 29) Huang KW, Ochandarena NE, Philson AC, Hyun M, Birnbaum JE, Cicconet M, Sabatini BL. Molecular and anatomical organization of the dorsal raphe nucleus. eLife, 8, e46464 (2019).
• 30) Ren J, Isakova A, Friedmann D, Zeng J, Grutzner SM, Pun A, Zhao GQ, Kolluru SS, Wang R, Lin R, Li P, Li A, Raymond JL, Luo Q, Luo M, Quake SR, Luo L. Single-cell transcriptomes and whole-brain projections of serotonin neurons in the mouse dorsal and median raphe nuclei. eLife, 8, e49424 (2019).
• 31) Paquelet GE, Carrion K, Lacefield CO, Zhou P, Hen R, Miller BR. Single-cell activity and network properties of dorsal raphe nucleus serotonin neurons during emotionally salient behaviors. Neuron, 110, 2664–2679.e8 (2022).
• 32) Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks. Commun. ACM, 60, 84–90 (2017).
• 33) Szegedy C, Ioffe S, Vanhoucke V, Alemi AA. Inception-v4, inception-ResNet and the impact of residual connections on learning. AAAI. 31st AAAI Conf. Artif. Intell. AAAI 2017, 31, 4278–4284 (2017).
• 34) Duvenaud D, Maclaurin D, Aguilera-Iparraguirre J, Gómez-Bombarelli R, Hirzel T, Aspuru-Guzik A, Adams RP. Convolutional networks on graphs for learning molecular fingerprints. Adv. Neural Inf. Process. Syst., 28, 2224–2232 (2015).
• 35) Wu Z, Ramsundar B, Feinberg EN, Gomes J, Geniesse C, Pappu AS, Leswing K, Pande V. MoleculeNet: a benchmark for molecular machine learning. Chem. Sci. (Camb.), 9, 513–530 (2018).
• 36) Mendez D, Gaulton A, Bento AP, et al. ChEMBL: towards direct deposition of bioassay data. Nucleic Acids Res., 47 (D1), D930–D940 (2019).
• 37) Cortés-Ciriano I, Bender A. KekuleScope: prediction of cancer cell line sensitivity and compound potency using convolutional neural networks trained on compound images. J. Cheminform., 11, 41 (2019).
• 38) Sakai M, Nagayasu K, Shibui N, Andoh C, Takayama K, Shirakawa H, Kaneko S. Prediction of pharmacological activities from chemical structures with graph convolutional neural networks. Sci. Rep., 11, 525 (2021).
• 39) Dash S, Shakyawar SK, Sharma M, Kaushik S. Big data in healthcare: management, analysis and future prospects. J. Big Data, 6, 54 (2019).
• 40) Zhao S, Nishimura T, Chen Y, Azeloglu EU, Gottesman O, Giannarelli C, Zafar MU, Benard L, Badimon JJ, Hajjar RJ, Goldfarb J, Iyengar R. Systems pharmacology of adverse event mitigation by drug combinations. Sci. Transl. Med., 5, 206ra140 (2013).
• 41) Nagashima T, Shirakawa H, Nakagawa T, Kaneko S. Prevention of antipsychotic-induced hyperglycaemia by vitamin D: A data mining prediction followed by experimental exploration of the molecular mechanism. Sci. Rep., 6, 26375 (2016).
• 42) Nagaoka K, Nagashima T, Asaoka N, Yamamoto H, Toda C, Kayanuma G, Siswanto S, Funahashi Y, Kuroda K, Kaibuchi K, Mori Y, Nagayasu K, Shirakawa H, Kaneko S. Striatal TRPV1 activation by acetaminophen ameliorates dopamine D2 receptor antagonist-induced orofacial dyskinesia. JCI Insight, 6, e145632 (2021).
• 43) Siswanto S, Yamamoto H, Furuta H, Kobayashi M, Nagashima T, Kayanuma G, Nagayasu K, Imai Y, Kaneko S. Drug repurposing prediction and validation from clinical big data for the effective treatment of interstitial lung disease. Front. Pharmacol., 12, 635293 (2021).
• 44) Furuta H, Yamada M, Nagashima T, Matsuda S, Nagayasu K, Shirakawa H, Kaneko S. Increased expression of glutathione peroxidase 3 prevents tendinopathy by suppressing oxidative stress. Front. Pharmacol., 14, 1137952 (2023).
• 45) Zhou Z, Nagashima T, Toda C, Kobayashi M, Suzuki T, Nagayasu K, Shirakawa H, Asai S, Kaneko S. Vitamin D supplementation is effective for olanzapine-induced dyslipidemia. Front. Pharmacol., 14, 1135516 (2023).
• 46) Yamamoto H, Kayanuma G, Nagashima T, Toda C, Nagayasu K, Kaneko S. Early detection of adverse drug reaction signals by association rule mining using large-scale administrative claims data. Drug Saf., 46, 371–389 (2023).