Comprehensive computational study of glutaminyl cyclase inhibitors, TRPV1 antagonists, and cytochrome P450 inhibitors

Abstract

Part I. Comprehensive computational study of human glutaminyl cyclase inhibitors for the drug discovery of Alzheimer’s disease Glutaminyl cyclase (QC) is an enzyme that catalyzes the N-terminal modification of glutamyl or glutaminyl precursors into pyroglutamate. The formation of pyroglutamate-modified Aβ peptide triggers the accumulation of Aβ oligomer and proto-fibrils causing neurotoxicity like brain cell death. Since these toxic materials aggravate neurodegenerative disorders like Alzheimer’s disease (AD), human QC (hQC) inhibitors could be a therapeutic target for AD. Recently, potent hQC inhibitors were designed and synthesized based on the pharmacophoric analysis of the substrate. Among the resulting compounds, two types of compounds showed excellent activities upon addition of functional groups relative to the previously reported inhibitors. In order to investigate the reason of the enhanced activities, docking study of these inhibitors using Glide SP docking was performed. For achieving more accurate docking results, QM-Polarized Ligand Docking and protein-ligand complex refinement using Local Optimization and Minimization Monte Carlo sampling algorithm were performed. These comprehensive analyses showed that the added functional groups in the first generation compounds could enable the formation of strong polar interactions with the side chain of Glu327, thereby playing a critical role in enhancing the inhibitory activity against hQC. On the other hand, despite the absence of the D-region seen in the first generation, second generation compounds still showed potent inhibitory activities against hQC may be attributed to additional hydrophobic interactions which may contribute to the conformation stability. Notably, one compound from second generation exhibited similar binding mode with the first generation compounds by forming polar interactions with the key residue, Glu327. Taken together, these computational studies could provide valuable insights into the development of promising drug candidates for halting or slowing the progression of AD. Part II. Analyses of TRPV1 structure and binding modes of its antagonists for neuropathic pain drug discovery Transient receptor potential (TRP) channels belong to a superfamily of sensory-related ion channels responding to a wide variety of thermal, mechanical or chemical stimuli. In an attempt to comprehend the piquancy and pain mechanism of the archetypal vanilloids transient receptor potential vanilloid (TRPV) 1 was discovered. TRPV1, a well-established member of the TRP family, is implicated in a range of functions including inflammation, pain stimuli sensation, mechanotransduction, and hyperalgesia. TRPV1 channels are non-selective cation receptors that are gated by a broad array of noxious ligands. Such polymodal-sensor aspect makes the TRPV1 channel extremely versatile and important for its role in sensing burning pain. Besides ligands, TRPV1 signaling can also be modulated by lipids, secondary messengers, protein kinases, cytoskeleton and several other proteins. Due to its central role in hyperalgesia transduction and inflammatory processes, it is considered as the primary pharmacological pain target. Moreover, understanding the structural and functional intricacies of the channel is indispensable for the therapeutic intervention of TRPV1 in pain and other pathological disorders. Additionally, the results about structure activity relationships between previously built TRPV1 homology model and the designed and synthesized antagonists were also proposed. The results of this study indicate the effective therapeutic possibility of the TRPV1. Part III. In silico classification of cytochrome P450 inhibitors using machine learning methods The Cytochrome P450 (CYP) enzyme superfamily is involved in oxidative phase I metabolism, which is a critical determinant for the metabolism of various xenobiotics, including drugs. Inhibition of CYPs by other drugs or chemicals can therefore lead to undesirable result, such as an increase of toxicity caused by a decreased drug metabolism rate or drug-drug interactions (DDI). Models that predict whether a compound interacts with a specific CYP isoform are therefore desirable to have. Five major CYP isoforms are closely associated with more than 80% of the metabolism of all the pharmaceuticals in clinical use. The purpose of this study is to develop in silico classification models that effectively distinguish inhibitors from non-inhibitors for each CYP isoform. Four machine learning methods, i.e., Laplacian-modified naïve Bayesian, random forest (RF), recursive partitioning (RP), and support vector machine (SVM) were utilized. These methods were selected due to their high speed and accuracy. Multiple-category modeling was performed, as well as single-category modeling. Data sets were collected from the public domain and randomly divided into training sets and test sets at a ratio of 7 to 3. The classification models were built using VolSurf+ descriptors and FCFP_8 fingerprint. The quality of each model was evaluated by its accuracy, sensitivity, and specificity with regard to the areas under the curves, based on the receiver operating characteristic (ROC) scores and Matthews correlation coefficient (MCC). The models constructed in this study could successfully classify inhibitors and non-inhibitors for each CYP isoform, and these models could be helpful in predicting the CYP inhibition profiles in early drug discovery.;Part I. Comprehensive Computational Study of Human Glutaminyl Cyclase Inhibitors for the Drug Discovery of Alzheimer’s Disease Glutaminyl cyclase (QC)는 N-terminal의 glutamyl 또는 glutaminyl을 pyroglutamate로 만드는 효소이다. 만들어진 pyroglutamate-modified Aβ 펩타이드는 Aβ 다량체로 응합되고, 이는 뇌세포 사멸 등 신경독성을 야기하여 신경퇴행성 질환인 알츠하이머를 악화시키는 등 매우 유해한 독성물질로 알려져 있다. 따라서 human QC(hQC)를 저해함으로써 알츠하이머 치료 효과를 발현할 수 있을 것으로 보고되고 있다. 최근들어 hQC 천연 기질의 약리작용단을 토대로 hQC 저해제들이 디자인 및 합성되었는데, 그 중에서 작용기가 덧붙여진 두 종류의 화합물들이 상당히 좋은 저해활성을 보여주고 있다. 저해활성을 보이는 제1, 2세대 화합물의 결합모드를 분석하고자 분자도킹 및 QM-Polarized 리간드 도킹을 수행하였다. 얻어진 단백질-리간드 복합체는 Local Optimization과 Monte Carlo minimization을 통해 refine하였다. 제1세대 화합물들은 Glu327의 곁사슬과의 강한 정전기적 상호작응을 통해 더 좋아진 저해활성을 보이는 것으로 나타났고, 제2세대 화합물들도 덧붙여진 작용기가 소수성 상호작용과 Glu327과의 정전기적 상호작응 등을 통해 개선된 저해효과를 나타내는 것으로 판단되었다. 이러한 결과들은 알츠하이머 치료제를 개발하는데 있어서 중요한 정보를 제공해줄 수 있을 것이다. Part II. Analyses of TRPV1 structure and binding modes of its antagonists for neuropathic pain drug discovery Transient receptor potential vanilloid subtype 1 (TPRV1)은 중추와 말초신경계에서 존재하는 비선택적 양이온 채널로서 열이나 기계적, 화학적 자극에 의해 활성화되며 염증 및 통각감지 등에 관여한다. 효과적인 통증치료제 개발을 위해 디자인 및 합성된 TRPV1 길항제인 2-(3-fluoro-4-methylsulfonylaminophenyl)propanamide 유도체들을 human TRPV1 상동수용체 모델을 이용하여 flexible한 분자도킹을 수행하였고 자세한 결합모드를 분석하였다. 활성이 높은 길항제들은 TRPV1의 결합 부위에 매우 잘 결합하였고 특히, 활성을 높이기 위해 B- 및 C-region을 치환한 화합물들의 경우, 그 치환기들이 TRPV1의 소수성 포켓 부분과 추가적인 소수성 결합을 했을 때 활성이 높아지는 것을 확인할 수 있었다. Part III. In silico classification of cytochrome P450 inhibitors using machine learning methods Cytochrome P450 (CYP)은 phase I 대사에서 산화작용을 하는 중요한 효소 superfamily이다. 특히, CYP1A2, 2C9, 2C19, 2D6와 3A4는 현재 임상에서 사용 중인 전체 약물 대사의 약 80% 이상에 관여하는 중요한 효소로 알려져있다. 약물을 병용투여하는 경우, 다른 약물의 대사를 저해하는 것은 큰 문제를 유발할 수 있으므로 약물개발 초기에 CYP 저해제들을 탐지하여 그 화합물이 약물개발 후반에서 fail될 가능성을 미리 예측할 수 있으며 비용과 시간의 손실을 줄일 수 있을 것이다. NIH에서는 17,000여개의 약물들에 대해 CYP isoform별 저해효과를 측정하였고 본 연구에서는 NIH의 big data를 이용하여 그 화합물들을 구조를 기반으로 VolSurf+ descriptor들과 Pipeline Pilot의 fingerprints descriptor들을 계산하였다. Laplacian-modified naïve Bayesian, random forest (RF), recursive partitioning (RP), support vector machine (SVM) 등 네 가지 머신러닝 방법들을 이용하여 isoform별 저해제/비저해제 분류모델을 수립하였다. 각각의 모델들은 정확도, 민감도, 특수성 및 ROC, MCC 등 방법으로 평가한 결과, 매우 높은 수준의 모델로 판명되었다. 또한, 수립된 모델들에 대해 validation data set으로 검증한 결과, 매우 신뢰도가 높은 예측 결과를 보여주었다.