Reflections on improving drug success rates with AIDD and CADD
-
摘要:
随着人工智能（AI）和计算科学的迅速发展，特别是人工智能药物设计（AIDD）与计算机辅助药物设计（CADD）技术的引入，自然语言处理、图像识别、深度学习和机器学习等多种技术为新药开发提供了革命性的新途径，大幅提升了研发流程的效率和成功率。在药物发现过程中，AI技术加速了药物靶点的识别、候选药物的筛选、药理评估及质量检验，有效降低了研发风险和成本。本文深入探讨AIDD和CADD技术在药物研发中的应用，分析它们在提升药物设计成功率和药物研发效率方面的思考与探索，并探讨这些技术的未来发展趋势及可能面临的挑战。
Abstract:The rapid advancements in artificial intelligence (AI) and computational sciences, particularly through the introduction of artificial intelligence drug design (AIDD) and computer-aided drug design (CADD) technologies, have revolutionized pathways in drug development. These include techniques such as natural language processing, image recognition, deep learning, and machine learning. By employing advanced algorithms and data processing techniques, these technologies have significantly enhanced the efficiency and success rate of R&D processes. In drug discovery, AI technologies have accelerated the identification of drug targets, screening of candidate drugs, pharmacological assessments, and quality control, effectively reducing R&D risks and costs. This article delves into the application of AIDD and CADD in drug development, analyzing their roles in enhancing the success rates and efficiencies of drug design, exploring their future trends, and addressing the potential challenges.
-
干扰素（interferon，IFN）是在免疫调节中具有多效性作用的细胞因子，根据序列同源性可分为3个家族（Ⅰ型、Ⅱ型和Ⅲ型）。在人体中，Ⅰ型干扰素（IFN-I）是最大的IFN家族，包括 IFN-α（13个亚型）、IFN-β、IFN-ε、IFN-κ和IFN-ω[1−2]。
IFN-I通过与其异二聚体型受体结合而启动信号，该受体被称为IFN-α/β受体（IFNAR）。IFNΑR是由两个跨膜蛋白IFNΑR1和IFNΑR2组成的异源二聚体，几乎在所有细胞类型上都有表达。人IFNΑR1属于Ⅱ型螺旋型细胞因子受体，包含4个Ⅲ型纤连蛋白结构域的胞外域，1个跨膜域和100个氨基酸残基的胞内域[3]。当IFN-I与IFNAR胞外域结合时，IFNAR1以及IFNAR2被活化，胞内区级联激活下游信号蛋白：活化的IFNAR1与酪氨酸激酶2(tyrosine kinase 2，Tyk2)结合，活化的IFNAR2与Janus激酶1(JAK1)结合，激活信号转导和转录激活因子（STATs）、丝裂原活化蛋白激酶（MAPK）和磷脂酰肌醇3-激酶（PI3K）信号通路以产生相应生物学效应[4]。
人IFN-α的异常表达与许多自身免疫性疾病相关，包括系统性红斑狼疮（SLE）、1型糖尿病、银屑病、风湿性关节炎、多发性硬化症、获得性免疫缺陷综合征和严重的混合免疫缺陷疾病等。大量数据显示，IFN-I是系统性红斑狼疮发病机制的重要参与者。据报道，60%~80%的系统性红斑狼疮患者存在IFN-I高表达特征，系统性红斑狼疮患者外周血细胞中IFN-I调控基因的过表达与系统性红斑狼疮疾病活动度正相关[5−6]。IFN-I与免疫激活的标志物（如补体）和自身抗体的产生（如抗 ds-DNA 抗体）相关，并且参与维持 SLE 疾病活动，提示IFN-I（特别是IFN-α）在SLE发病机制中的重要作用[7]。因此，阻断IFN-I信号通路对SLE具有潜在的治疗作用。
针对IFN-α开发的单克隆抗体有阿斯利康的西法木单抗（sifalimumab）和罗氏的隆利组单抗（rontalizumab）。隆利组单抗的Ⅱ期临床试验未达到主要终点BILAG指数的改善，已停止开发[8]。虽然西法木单抗的Ⅱ期临床达到主要终点，但是阿尼鲁单抗（anifrolumab）具有强劲的药效学反应，加上比西法木单抗更好的获益风险比，使阿尼鲁单抗成为系统性红斑狼疮Ⅲ期开发的候选药物[9−10]。阿尼鲁单抗通过直接作用于IFNAR1，能够阻断与SLE发病机制有关的几种IFN-I（IFN-α、IFN-β 和 IFN-ω）与其受体的结合，抑制IFN-I的信号转导及生物活性。与安慰剂相比，更多接受阿尼鲁单抗治疗的患者，整个器官系统（包括皮肤和关节）的总体疾病活动度降低，并且口服皮质类固醇的使用持续减少。2021年7月，阿尼鲁单抗被美国FDA批准，用于治疗正在接受标准治疗的中度至重度系统性红斑狼疮成人患者[11−12]。
目前临床上治疗性单克隆抗体多为鼠源，相比之下，基于B细胞克隆技术的兔单克隆抗体平台更易获得多样性佳、亲和力高、功能性优、特异性强的单克隆抗体，逐渐被应用和重视。本研究采用人IFNAR1蛋白免疫新西兰白兔，利用B细胞克隆技术制备兔单抗；筛选出IFN-I信号中和活性最好的亲本抗体，通过互补决定区（CDR）移植进行人源化改造获得QX006N；经过体外生物学活性评价，显示出优良的生物学特性。该研究为靶向IFNAR1抗体药物及其临床试验开展奠定了基础。
1. 材 料
1.1 试 剂
人IFNAR1、人IFNAR2、人IFNGR1、人IFN-α2b和人IFN-γ（苏州近岸蛋白质科技股份有限公司）；人IFN-β和人IFN-ω1（北京义翘神州科技股份有限公司）；人IFNGR2（美国Novus Biologicals公司）；人IFN-ε（美国R&D公司）。根据专利WO2009100309A2提供的9D4序列，构建表达质粒，瞬转ExpiCHO-S细胞自制获得阿尼鲁单抗。
1.2 仪 器
Varioskan LUX多功能酶标仪，NanoDrop One超微量分光光度计（美国 Thermo Fisher Scientific公司）；Biacore T200生物分子相互作用分析仪（美国GE公司）。
1.3 细胞株
ExpiCHO-S细胞（美国 Thermo Fisher Scientific公司）；HEK Blue™ IFN-α/β细胞（美国InvivoGen公司）；Daudi细胞和THP-1细胞（上海盖宁生物科技有限公司）。根据《赫尔辛基宣言》，于健康献血者手臂抽取外周血（肝素钠抗凝）。
2. 方 法
2.1 兔抗人IFNAR1单克隆抗体的制备
采用人IFNAR1重组蛋白免疫新西兰白兔；提取经免疫动物外周血单个核细胞，通过B细胞克隆技术分离培养抗原特异性B细胞；利用ELISA及HEK BlueTM IFN-α/β报告基因细胞法分析B细胞克隆上清液的结合和中和活性。提取有中和活性的B细胞克隆的mRNA，通过RT-PCR获取抗体的基因序列。将抗体重链和轻链可变区分别与载体pRBT进行构建，瞬转ExpiCHO-S细胞进行抗体的表达，用ProteinA亲和色谱法纯化抗体，经HEK BlueTM IFN-α/β报告基因细胞法筛选出中和活性最佳的亲本兔抗用于人源化改造。
2.2 抗人IFNAR1兔单克隆抗体的人源化改造
经筛选，362#兔抗和1203#兔抗的中和活性最好，序列高度相似。利用NCBI IgBlast（https://www.ncbi.nlm.nih.gov/projects/igblast/）进行人IgG胚系序列（germline）同源性比对，选择同源性最高的IGHV3-66*01作为重链CDR移植模板，将兔抗重链的CDR区移植入IGHV3-66*01的骨架区；选择同源性最高的IGKV1-6*01作为轻链CDR移植模板，将兔抗轻链的CDR区移植入IGKV1-6*01的骨架区；对骨架区特定位点进行回复突变，获得人源化抗体的可变区。为了减少ADCC效应带来的潜在的不良反应，人源化抗体重链恒定区选择人IgG4亚型。重组人源化抗体序列重链和轻链分别与载体pHZD进行构建，并表达抗体，方法同“2.1”项。利用分子排阻色谱高效液相（SEC-HPLC）检测抗体纯度，并用HEK BlueTM IFN-α/β报告基因细胞法对抗体分子的中和能力进行比较，筛选中和能力强的抗体分子作为候选抗体进行进一步评价。
2.3 ELISA检测QX006N与抗原结合的活性及特异性
分别将IFNAR1以及相关蛋白（人IFNAR2、人IFNGR1、人IFNGR2、人IFN-α2b、人IFN-β、人IFN-γ、人IFN-ω1、人IFN-ε）以1 μg/mL（每孔50 μL）包被酶标板，2～8 ℃过夜后，弃去包被液，洗板，加0.5% BSA-PBS封闭2 h。加入10 μg/mL起始1∶5梯度稀释的QX006N，每孔50 μL，置于室温孵育2 h。洗板后每孔加入 100 ng/mL HRP标记的羊抗人IgG抗体50 μL置于室温孵育1 h，洗板后加入TMB进行显色反应，用酶标仪在450、630 nm处读取吸收度，并采用SoftMax软件进行四参数拟合，分析评价QX006N的抗原结合活性和特异性。
2.4 表面等离子共振技术（SPR）检测QX006N的亲和力
用Biacore T200检测QX006N与人IFNAR1的亲和力，所有过程于25 ℃进行。采用商品化Protein A芯片，通过捕获法固定适量的QX006N，捕获流速是10 μL/min。将抗原进行梯度稀释，仪器流速切换成30 μL/min，按照浓度（20 nmol/L，1∶2梯度稀释）从低到高的顺序依次流过参比通道和固定抗体的通道，流过缓冲液作为阴性对照。每个结合、解离完成后用pH 1.5甘氨酸再生芯片。用仪器自带分析软件选择Kinetics选项中1∶1结合模型进行拟合，计算抗体的结合速率常数ka，解离速率常数kd以及解离平衡常数KD（kd/ka）。
2.5 细胞水平检测QX006N的中和活性
报告基因细胞是检测信号通路激活与抑制常用的细胞模型。IFN-I可诱导 HEK Blue™ IFN-α/β报告细胞STAT1/2磷酸化，抗体中和IFN-I能力越强，报告细胞 STAT1/2磷酸化信号则越弱。此外，由于IFN-I可与多种细胞增殖或下游因子释放有关，本研究将分别以Daudi细胞、THP-1细胞和人全血细胞3种不同的细胞为模型检测抗体的中和能力。相关检测方法如表1所示。采用SoftMax拟合四参数曲线，计算样品IC50，分析评价QX006N的中和活性。
Table 1. Different neutralization assays to evaluate the potency of QX006NAssay Stimulating factor c/(ng/mL) Detection method HEK BlueTM
IFN-α/β cellIFN-α2b 0.05 STAT1/2 phosphorylation IFN-β 0.005 IFN-ω1 1 Daudi cell IFN-α2b 0.2 Cell proliferation THP-1 cell IFN-α2b 10 Release of IP-10 and BLyS Human whole blood cell IFN-α2b* 1 Release of IP-10 * with 5 ng/mL of TNFα
IP-10: Interferon gamma-induced protein 10; BLyS: B lymphocyte stimulator protein2.6 ELISA分析QX006N的抗原结合表位区域
人IFNAR1的胞外区（hIFNAR1-ECD）包含4个结构域Domain1（D1，K28~A126）、Domain2（D2，Q127~N227）、Domain3（D3，E228~Q329）和Domain4（D4，A330~K436），分别对应图1中蓝色、绿色、黄色和灰色背景的氨基酸序列。根据hIFNAR1-ECD的序列特征，设计并制备截短的hIFNAR1-ECD突变体，包括hIFNAR1（D1+D2）、hIFNAR1（D3+D4）和hIFNAR1（D1+D2+D3），通过ELISA检测突变体与QX006N的结合情况，分析 QX006N与IFNAR1-ECD的结合区域位置。
![]()
Figure 1. Human IFNAR1-ECD amino acid sequenceUnderlined sequence is the signal peptide sequence, and the amino acid sequences with blue, green, yellow and gray backgrounds constitute Domain1 to Domain4 of IFNAR1-ECD, respectively3. 结 果
3.1 兔抗人IFNAR1单克隆抗体的制备和筛选
本研究共进行4轮B细胞克隆和筛选，获得114个具有中和活性的B细胞克隆；优选13个克隆进行重组表达后进行中和活性验证，最终挑选2个中和活性最佳的兔单克隆抗体362#和1203#（图2）。兔抗362#和1203#的氨基酸序列高度相似，分别在重链CDR2区和轻链CDR2区相差1个氨基酸（图3）。HEK BlueTM IFN-α/β报告基因细胞活性检测结果（表2）显示，兔抗1203#的中和活性更优。
![]()
Figure 2. Neutralizing activity of 13 recombinant rabbit monoclonal antibodies compared to anifrolumab![]()
Figure 3. Amino acid sequence alignment between the variable region of rabbit mAb 362# and 1203#CDR region of the antibody is marked with an underscore according to the Kabat numbering scheme; “-” means the amino acid of 1203# is identical to 362#; VH：Heavy-chain variable region; VL：Light-chain variable regionTable 2. Neutralization activity of rabbit mAb 362# and 1203# based on HEK-BlueTM IFN-α/β reporter cellsSample IC50/(ng/mL) IFN-α2b IFN-β IFN-ω1 Anifrolumab 5.5 28.3 23.3 362# 5.8 27.2 29.1 1203# 4.7 23.4 17.5 3.2 抗人IFNAR1兔单克隆抗体的人源化
通过CDR移植结合骨架区关键氨基酸回复突变，对362#兔抗进行人源化改造（图4）。结果（表3）显示，与对照阿尼鲁单抗相比，人源化改造导致中和活性下降，HZD362-5的相对中和活性只有72.3%。由于1203#兔抗的细胞活性相比362#略好，且1203#兔抗的CDR区比362#兔抗多了2个与人germline相似的氨基酸，因此继续对1203#兔抗进行人源化改造（图4）。与对照阿尼鲁单抗相比，HZD1203-45的细胞比活达92.0%，选为最终的人源化分子QX006N。
![]()
Figure 4. Amino acid sequence alignment between the variable regions of humanized anti-IFNAR1 rabbit monoclonal antibodies“-” denotes residues that are identical to human germline at the corresponding positions; “#” denotes the residues in human framework regions were back-mutated; “*” denotes the different residues between rabbit antibodies 362# and 1203#Table 3. Summary of humanized anti-IFNAR1 rabbit monoclonal antibodiesAntibody VH VL SEC-HPLC/% Activity ratio* Name Humanization Name Humanization HMW Monomer LMW HZD362-1 362VH-Hu1 97.7% 362VK-Hu1 100% 2.7 97.2 0.1 21.2% HZD362-5 362VH-Hu1 97.7% 362VK-Hu2 93.8% 1.2 98.8 ND 72.3% HZD1203-38 362VH-Hu6 97.7% 362VK-Hu13 98.8% 0.8 99.2 ND 35.0% HZD1203-39 362VH-Hu6 97.7% 362VK-Hu14 98.8% 0.8 99.2 ND 34.0% HZD1203-45 362VH-Hu6 97.7% 362VK-Hu20 97.5% 1.3 98.7 ND 92.0% ND:Not detected; HMW:High molecular weight; LMW:Low molecular weight
* Neutralization activity of humanized antibodies compared to anifrolumab was measured using HEK Blue™ IFN-α/β reporter cell3.3 QX006N的抗原结合活性和特异性
本研究运用ELISA的方法，检测IFNAR1以及相关蛋白（人IFNAR2、人IFNGR1、人IFNGR2、人IFN-α2b、人IFN-β、人IFN-γ、人IFN-ω1、人IFN-ε）的结合情况。检测结果显示，QX006N与人IFNAR1特异性结合，与相关蛋白均无结合（图5）。
![]()
Figure 5. Binding map of QX006N to related proteinsA: hIFNAR2 and hIFNGR1; B: hIFNGR2 and hIFN-α2b; C: hIFN-β and hIFN-γ; D: hIFN-ε and hIFN-ω13.4 QX006N的抗原亲和力
采用Biacore T200检测QX006N、阿尼鲁单抗与人IFNAR1的亲和力，结果显示QX006N与人IFNAR1的亲和力常数KD与阿尼鲁单抗相当（表4）。
Table 4. Affinity of QX006N and anifrolumab binding to human IFNAR1 ($ \bar{x} $± s, n = 3)Sample ka/(×105 L∙mol-1∙s-1) kd/(×10-5/s) KD/(×10-10 mol/L) QX006N 3.47 3.76 1.08 Anifrolumab 18.67 12.40 0.67 3.5 QX006N的中和活性
3.5.1 HEK Blue™ IFN-α/β细胞法
采用HEK Blue™ IFN-α/β报告基因细胞法分析QX006N对IFN-I（IFN-α2b、IFN-β和IFN-ω1）诱导HEK Blue™ IFN-α/β细胞STAT1/2磷酸化活性的中和作用。结果显示（图6、表5）：QX006N中和IFN-α2b、IFN-β、IFN-ω1诱导的报告基因细胞STAT1/2磷酸化活性的IC50分别为（4.6 ± 0.4）、（17.4 ± 1.2）和（12.1 ± 1.0）ng/mL，与阿尼鲁单抗相当。
![]()
Figure 6. QX006N inhibits different IFN-I induced STAT1/2 phosphorylation in HEK Blue™ IFN-α/β cells ($ \bar{x} $± s, n = 3)A: IFN-α2b; B: IFN-β; C: IFN-ω1Table 5. Neutralization activity of QX006N based on HEK Blue™ IFN-α/β cell ($ \bar{x} $± s, n = 3)Sample IC50/(ng/mL) IFN-α2b IFN-β IFN-ω1 QX006N 4.6 ± 0.4 17.4 ± 1.2 12.1 ± 1.0 Anifrolumab 4.0 ± 0.2 17.4 ± 1.3 10.6 ± 0.7 IgG4 Isotype No neutralizing activity 3.5.2 Daudi细胞法
采用Daudi细胞法分析QX006N中和IFN-α2b抑制Daudi细胞增殖的活性。结果显示（图7、表6）：QX006N能够中和IFN-α2b诱导的Daudi细胞增殖，其IC50为（31.4 ± 1.6）ng/mL，与阿尼鲁单抗活性相当。
![]()
Figure 7. QX006N inhibits IFN-α2b induced cell proliferation and cytokine release ($ \bar{x} $± s, n = 3)A: Daudi cell proliferation; B: IP-10 release in THP-1 cell; C: BLyS release in THP-1 cell; D: IP-10 release in human whole blood cellTable 6. Neutralization activity of QX006N based on Daudi, THP-1 and human whole blood cells ($ \bar{x} $±s, n = 3)Sample IC50/(ng/mL) Daudi cells proliferation THP-1 cells IP-10 release THP-1 cells BLyS release Blood cells IP-10 release QX006N 31.4 ± 1.6 2.0 ± 0.5 5.9 ± 1.4 1003 ± 311 anifrolumab 31.3 ± 1.7 7.1 ± 0.8 101.3 ± 28.2 904 ± 314 IgG4 Isotype No neutralizing activity 3.5.3 THP-1细胞法
采用THP-1细胞法分析QX006N中和IFN-α2b诱导THP-1细胞释放IP-10和BLyS的活性。结果显示（图7、表6）：QX006N能够中和IFN-α2b诱导THP-1细胞释放IP-10和BLyS，其IC50分别为（2.0 ± 0.5）ng/mL和（5.9 ± 1.4）ng/mL，活性优于阿尼鲁单抗。
3.5.4 人全血法
采用人全血法分析QX006N中和IFN-α2b诱导人全血释放IP-10的活性。结果显示（图7、表6）：QX006N能够中和IFN-α2b诱导人全血释放IP-10，其IC50为（1003 ± 311）ng/mL，与阿尼鲁单抗活性相当。
3.6 QX006N的抗原结合表位区域
QX006N是靶向IFNAR1胞外区的人源化单克隆抗体。人IFNAR1（NP_000620.2）是含有557个氨基酸的膜受体蛋白，其中1~27位氨基酸序列是IFNAR1的信号肽，28~436位氨基酸序列是IFNAR1的胞外区(IFNAR1-ECD)，胞外区共包含4个区域（图1）。ELISA检测QX006N与截短的hIFNAR1突变体EC50显示（表7），hIFNAR1（D3+D4）和hIFNAR1（D1+D2+D3）与QX006N的EC50与天然hFNAR1-ECD基本一致，而hIFNAR1（D1+D2）不与QX006N结合，因此推测QX006N与hIFNAR1-ECD的结合在Domain3。
Table 7. QX006N binds to truncated hIFNAR1-ECD mutantsSample QX006N EC50/(ng/mL) Ratio hIFNAR1-ECD 8.26 1.0 hIFNAR1(D1+D2) NA NA hIFNAR1(D3+D4) 4.29 0.5 hIFNAR1(D1+D2+D3) 8.53 1.0 NA: No binding 4. 讨 论
IFN-I结合IFNAR1和IFNAR2形成三元复合物，激活下游JAK-STAT信号通路。为了阻断IFN-I信号通路，抗体可以结合IFNAR1，也可以结合IFNAR2。据报道，所有 IFN均以微摩尔亲和力结合 IFNAR1，并以纳摩尔亲和力结合 IFNAR2[13]。抗体结合IFNAR1并阻断下游信号通路，比结合IFNAR2更容易，因此选择IFNAR1作为靶点开发治疗药物。
兔比小鼠具有更大的体型，有助于收集更多的免疫B细胞，产生更多样化的兔单抗[14]。兔的 B 细胞除发生体细胞超变，还存在基因转换的现象。基因转换正是兔免疫发育过程中不同于小鼠的关键特征，将会极大提升初级和次级B细胞库重链和轻链可变区的多样性和亲和力，为抗体的筛选提供更广泛的选择。诺华公司的布洛赛珠单抗（brolucizumab）和灵北公司的艾普奈珠单抗（eptinezumab）先后被美国食品药品监督管理局（FDA）批准上市[15−16]，充分证明了人源化兔单抗的成药性。因此，通过免疫兔筛选获得的兔单抗，具备多样性好、亲和力高以及成药效佳等潜在优势。
有研究表明人源化改造有可能导致抗体的活性下降。本研究发现对筛选获得的兔抗进行人源化改造后，抗体活性下降。其中HZD362-5的细胞比活只有72.3%，且HZD362-5轻链骨架区有5个位点进行了回复突变，导致轻链骨架区的人源化程度只有93.8%。由于1203#兔抗的细胞活性略好，且CDR区比362#兔抗的人源化程度更高，因此选择对1203#兔抗继续进行人源化改造。最终获得了人源化抗体QX006N（HZD1203-45），细胞活性约92%，与对照阿尼鲁单抗相当，且轻链骨架区的人源化程度提升到97.5%。
初步体外评价结果显示QX006N特异性结合人IFNAR1 Domain3区域，能够有效中和IFN-I（IFN-α2b、IFN-β和IFN-ω1）诱导HEK Blue™ IFN-α/β细胞STAT1/2磷酸化作用，表明QX006N能有效中和IFN-I信号通路，可作为Ⅰ型IFN信号通路的有效拮抗剂。而且，QX006N能够分别有效中和IFN-α2b诱导的Daudi细胞增殖，THP-1细胞释放IP-10和BLyS，以及人全血释放IP-10。以上结果表明QX006N能够有效中和IFN-I介导的细胞下游生物学效应，为其在SLE治疗中的临床应用提供了坚实的基础。目前，QX006N正在进行进一步的临床试验，旨在验证其安全性、有效性以及在SLE治疗中的潜在优势。该研究不仅为IFN-I信号在SLE等自身免疫疾病中的角色提供了更深入的理解，也展示了基于兔抗体平台开发的单克隆抗体在治疗这类疾病中的巨大潜力。
-
表 1 药效团中药效特征元素[18]
特征元素 药效团内容 氢键受体 sp2或sp3杂化的氧原子/与碳原子以双键形式相连的S原子/与碳原子以双键或者三键相连的氮原子 氢键供体 氢原子以及与之相连的氧原子和氮原子 疏水中心 只要和不带电原子或电负性中心相连的一组连续的碳原子都可以形成疏水中心 电荷中心 与受体形成盐桥或较强的静电相互作用，带有电荷的原子/在生理pH下会发生电离的中性基团 芳环中心 形成π-π相互作用，五元或六元芳环 
下载: 导出CSV
表 2 按照药物设计理念分类的分子生成方法
种 类 分子生成方法 经典方法 基于规则的方法 递归神经网络 生成对抗网络 变分自编码器 强化学习 融入药学思想的生成模型 基于受体结构的生成模型 基于配体结构的生成模型 基于药效团的生成模型 基于片段的生成模型 其他 基于谱学的生成模型 基于化学反应的生成模型
下载: 导出CSV
表 3 经筛选得到的5-HT1A受体小分子活性数据
Hit 相对分子质量 氢键受体 氢键供体 ALogP [3H]-8-OH-DPAT Ki/(nmol/L) [35S]GTPyS EC50/(nmol/L) FW01 476.6 5 1 5.84 51.9±16.4 7 FW02 347.5 4 2 3.35 70.5±14.8 77 FW03 350.4 4 1 4.04 96.5±13.1 404 FW04 394.5 5 2 3.77 103.5±23.4 128 FW05 429.0 4 1 6.10 123.6±32.0 534 FW06 294.3 4 3 3.27 133.4±9.7 434 FW07 397.5 4 1 4.60 225.2±20.2 340 FW08 344.8 4 2 4.20 294.9±5.0 572 FW09 446.0 5 1 4.74 320.1±76.9 597 FW10 396.9 4 1 4.27 351.7±19.0 ND 5-羟色胺 1.8±0.1 3
下载: 导出CSV
-
[1] DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: new estimates of R&D costs[J]. J Health Econ, 2016, 47: 20-33. doi: 10.1016/j.jhealeco.2016.01.012
[2] DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: new estimates of R&D costs[J]. J Health Econ, 2016, 47: 20-33. doi: 10.1016/j.jhealeco.2016.01.012
[3] Vemula D, Jayasurya P, Sushmitha V, et al. CADD, AI and ML in drug discovery: a comprehensive review[J]. Eur J Pharm Sci, 2023, 181: 106324. doi: 10.1016/j.ejps.2022.106324
[4] National Medical Products Administration. Novel Drug Screening R&D Platform-CADD[EB/OL].(2020-08-13)[2023-11-23]. http://www.cnpharm.com/c/2020-08-13/748891.shtmL. National Medical Products Administration. Novel Drug Screening R&D Platform-CADD[EB/OL].(2020-08-13)[2023-11-23]. http://www.cnpharm.com/c/2020-08-13/748891.shtmL.
[5] Liu XW, Shi DF, Zhou SY, et al. Molecular dynamics simulations and novel drug discovery[J]. Expert Opin Drug Discov, 2018, 13(1): 23-37. doi: 10.1080/17460441.2018.1403419
[6] Yan T, Yu L, Zhang N, et al. The advanced development of molecular targeted therapy for hepatocellular carcinoma[J]. Cancer Biol Med, 2022, 19(6): 802-817.
[7] Boomsma W, Nielsen SV, Lindorff-Larsen K, et al. Bioinformatics analysis identifies several intrinsically disordered human E3 ubiquitin-protein ligases[J]. PeerJ, 2016, 4: e1725. doi: 10.7717/peerj.1725
[8] Huang HJ, Yu HW, Chen CY, et al. Current developments of computer-aided drug design[J]. J Taiwan Inst Chem Eng, 2010, 41(6): 623-635. doi: 10.1016/j.jtice.2010.03.017
[9] Matsuzaka Y, Uesawa Y. Ensemble learning, deep learning-based and molecular descriptor-based quantitative structure-activity relationships[J]. Molecules, 2023, 28(5): 2410. doi: 10.3390/molecules28052410
[10] Bissaro M, Sturlese M, Moro S. The rise of molecular simulations in fragment-based drug design (FBDD): an overview[J]. Drug Discov Today, 2020, 25(9): 1693-1701. doi: 10.1016/j.drudis.2020.06.023
[11] Kashyap A, Singh PK, Silakari O. Counting on fragment based drug design approach for drug discovery[J]. Curr Top Med Chem, 2019, 18(27): 2284-2293. doi: 10.2174/1568026619666181130134250
[12] Sheng CQ, Zhang WN. Fragment informatics and computational fragment-based drug design: an overview and update[J]. Med Res Rev, 2013, 33(3): 554-598. doi: 10.1002/med.21255
[13] Konteatis ZD. In silico fragment-based drug design[J]. Expert Opin Drug Discov, 2010, 5(11): 1047-1065. doi: 10.1517/17460441.2010.523697
[14] Liu CF, Hu LD, Dong GY, et al. Emerging drug design strategies in anti-influenza drug discovery[J]. Acta Pharm Sin B, 2023, 13(12): 4715-4732. doi: 10.1016/j.apsb.2023.08.010
[15] Li QX, Kang CB. Perspectives on fragment-based drug discovery: a strategy applicable to diverse targets[J]. Curr Top Med Chem, 2021, 21(13): 1099-1112. doi: 10.2174/1568026621666210804115700
[16] Giordano D, Biancaniello C, Argenio MA, et al. Drug design by pharmacophore and virtual screening approach[J]. Pharmaceuticals, 2022, 15(5): 646. doi: 10.3390/ph15050646
[17] Dror O, Schneidman-Duhovny D, Inbar Y, et al. Novel approach for efficient pharmacophore-based virtual screening: method and applications[J]. J Chem Inf Model, 2009, 49(10): 2333-2343. doi: 10.1021/ci900263d
[18] Yang SY. Pharmacophore modeling and applications in drug discovery: challenges and recent advances[J]. Drug Discov Today, 2010, 15(11/12): 444-450.
[19] Zhong FS, Xing J, Li XT, et al. Artificial intelligence in drug design[J]. Sci China Life Sci, 2018, 61(10): 1191-1204. doi: 10.1007/s11427-018-9342-2
[20] Yang X, Wang YF, Byrne R, et al. Concepts of artificial intelligence for computer-assisted drug discovery[J]. Chem Rev, 2019, 119(18): 10520-10594. doi: 10.1021/acs.chemrev.8b00728
[21] Sarkar C, Das B, Rawat VS, et al. Artificial intelligence and machine learning technology driven modern drug discovery and development[J]. Int J Mol Sci, 2023, 24(3): 2026. doi: 10.3390/ijms24032026
[22] Jiménez-Luna J, Grisoni F, Weskamp N, et al. Artificial intelligence in drug discovery: recent advances and future perspectives[J]. Expert Opin Drug Discov, 2021, 16(9): 949-959. doi: 10.1080/17460441.2021.1909567
[23] Li Z, Jiang MJ, Wang S, et al. Deep learning methods for molecular representation and property prediction[J]. Drug Discov Today, 2022, 27(12): 103373. doi: 10.1016/j.drudis.2022.103373
[24] Lin CX, Guan YF, Li HD. Artificial intelligence approaches for molecular representation in drug response prediction[J]. Curr Opin Struct Biol, 2024, 84: 102747. doi: 10.1016/j.sbi.2023.102747
[25] Chuang KV, Gunsalus LM, Keiser MJ. Learning molecular representations for medicinal chemistry[J]. J Med Chem, 2020, 63(16): 8705-8722. doi: 10.1021/acs.jmedchem.0c00385
[26] Chen BY, Pan ZQ, Mou MJ, et al. Is fragment-based graph a better graph-based molecular representation for drug design? A comparison study of graph-based models[J]. Comput Biol Med, 2024, 169: 107811. doi: 10.1016/j.compbiomed.2023.107811
[27] Xiong ZP, Wang DY, Liu XH, et al. Pushing the boundaries of molecular representation for drug discovery with the graph attention mechanism[J]. J Med Chem, 2020, 63(16): 8749-8760. doi: 10.1021/acs.jmedchem.9b00959
[28] Marín M, López M, Gallego-Yerga L, et al. Experimental structure based drug design (SBDD) applications for anti-leishmanial drugs: a paradigm shift[J]? Med Res Rev, 2024, 44(3): 1055-1120. doi: 10.1002/med.22005
[29] Goodsell DS, Morris GM, Olson AJ. Automated docking of flexible ligands: applications of AutoDock[J]. J Mol Recognit, 1996, 9(1): 1-5. doi: 10.1002/(SICI)1099-1352(199601)9:1<1::AID-JMR241>3.0.CO;2-6
[30] Eberhardt J, Santos-Martins D, Tillack AF, et al. AutoDock vina 1.2. 0: new docking methods, expanded force field, and python bindings[J]. J Chem Inf Model, 2021, 61(8): 3891-3898. doi: 10.1021/acs.jcim.1c00203
[31] Wu YJ, Brooks III CL. Covalent docking in CDOCKER[J]. J Comput Aided Mol Des, 2022, 36(8): 563-574. doi: 10.1007/s10822-022-00472-3
[32] Friesner RA, Banks JL, Murphy RB, et al. Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy[J]. J Med Chem, 2004, 47(7): 1739-1749. doi: 10.1021/jm0306430
[33] Miyamoto Y, Yamauchi J, Sanbe A, et al. Dock6, a Dock-C subfamily guanine nucleotide exchanger, has the dual specificity for Rac1 and Cdc42 and regulates neurite outgrowth[J]. Exp Cell Res, 2007, 313(4): 791-804. doi: 10.1016/j.yexcr.2006.11.017
[34] David L, Mdahoma A, Singh N, et al. A toolkit for covalent docking with GOLD: from automated ligand preparation with KNIME to bound protein-ligand complexes[J]. Bioinform Adv, 2022, 2(1): vbac090. doi: 10.1093/bioadv/vbac090
[35] Cross SSJ. Improved FlexX docking using FlexS-determined base fragment placement[J]. J Chem Inf Model, 2005, 45(4): 993-1001. doi: 10.1021/ci050026f
[36] Grosdidier A, Zoete V, Michielin O. SwissDock, a protein-small molecule docking web service based on EADock DSS[J]. Nucleic Acids Res, 2011, 39(Web Server issue): W270-W277.
[37] Kaur T, Madgulkar A, Bhalekar M, et al. Molecular docking in formulation and development[J]. Curr Drug Discov Technol, 2019, 16(1): 30-39. doi: 10.2174/1570163815666180219112421
[38] Huang YP, Zhang H, Jiang SY, et al. DSDP: a blind docking strategy accelerated by GPUs[J]. J Chem Inf Model, 2023, 63(14): 4355-4363. doi: 10.1021/acs.jcim.3c00519
[39] B Fortela DL, Mikolajczyk AP, Carnes MR, et al. Predicting molecular docking of per- and polyfluoroalkyl substancesto blood protein using generative artificial intelligence algorithm DiffDock[J]. Biotechniques, 2024, 76(1): 14-26. doi: 10.2144/btn-2023-0070
[40] Ryazantsev MN, Nikolaev DM, Struts AV, et al. Quantum mechanical and molecular mechanics modeling of membrane-embedded rhodopsins[J]. J Membr Biol, 2019, 252(4/5): 425-449.
[41] Ding Y, Yu K, Huang J. Data science techniques in biomolecular force field development[J]. Curr Opin Struct Biol, 2023, 78: 102502. doi: 10.1016/j.sbi.2022.102502
[42] Hollingsworth SA, Dror RO. Molecular dynamics simulation for all[J]. Neuron, 2018, 99(6): 1129-1143. doi: 10.1016/j.neuron.2018.08.011
[43] Dheyab MA, Aziz AA, Rahman AA, et al. Monte Carlo simulation of gold nanoparticles for X-ray enhancement application[J]. Biochim Biophys Acta Gen Subj, 2023, 1867(4): 130318. doi: 10.1016/j.bbagen.2023.130318
[44] Ruff KM, Pappu RV. AlphaFold and implications for intrinsically disordered proteins[J]. J Mol Biol, 2021, 433(20): 167208. doi: 10.1016/j.jmb.2021.167208
[45] Loman TE, Ma YB, Ilin V, et al. Catalyst: fast and flexible modeling of reaction networks[J]. PLoS Comput Biol, 2023, 19(10): e1011530. doi: 10.1371/journal.pcbi.1011530
[46] Wolber G, Langer T. LigandScout: 3-D pharmacophores derived from protein-bound ligands and their use as virtual screening filters[J]. J Chem Inf Model, 2005, 45(1): 160-169. doi: 10.1021/ci049885e
[47] Benford D, Leblanc JC, Setzer RW. Application of the margin of exposure (MoE) approach to substances in food that are genotoxic and carcinogenic: example: aflatoxin B1 (AFB1)[J]. Food Chem Toxicol, 2010, 48(Suppl 1): S34-S41.
[48] Liu XF, Ouyang SS, Yu B, et al. PharmMapper server: a web server for potential drug target identification using pharmacophore mapping approach[J]. Nucleic Acids Res, 2010, 38(Web Server issue): W609-W614.
[49] Schneidman-Duhovny D, Dror O, Inbar Y, et al. PharmaGist: a webserver for ligand-based pharmacophore detection[J]. Nucleic Acids Res, 2008, 36(Web Server issue): W223-W228.
[50] Verma J, Khedkar VM, Coutinho EC. 3D-QSAR in drug design: a review[J]. Curr Top Med Chem, 2010, 10(1): 95-115. doi: 10.2174/156802610790232260
[51] Wang YL, Wang F, Shi XX, et al. Cloud 3D-QSAR: a web tool for the development of quantitative structure-activity relationship models in drug discovery[J]. Brief Bioinform, 2021, 22(4): bbaa276. doi: 10.1093/bib/bbaa276
[52] Ataide Martins JP, Rougeth de Oliveira MA, Oliveira de Queiroz MS. Web-4D-QSAR: a web-based application to generate 4D-QSAR descriptors[J]. J Comput Chem, 2018, 39(15): 917-924. doi: 10.1002/jcc.25166
[53] Zhu C, Li XW, Zhao BY, et al. Discovery of aryl-piperidine derivatives as potential antipsychotic agents using molecular hybridization strategy[J]. Eur J Med Chem, 2020, 193: 112214. doi: 10.1016/j.ejmech.2020.112214
[54] Feng W, Wang LW, Lin ZY, et al. Generation of 3D molecules in pockets via a language model[J]. Nat Mach Intell, 2024, 6: 62-73. doi: 10.1038/s42256-023-00775-6
[55] Qian H, Lin C, Zhao DW, et al. AlphaDrug: protein target specific de novo molecular generation[J]. PNAS Nexus, 2022, 1(4): pgac227. doi: 10.1093/pnasnexus/pgac227
[56] Kosugi T, Ohue M. Quantitative estimate index for early-stage screening of compounds targeting protein-protein interactions[J]. Int J Mol Sci, 2021, 22(20): 10925. doi: 10.3390/ijms222010925
[57] Creanza TM, Lamanna G, Delre P, et al. DeLA-drug: a deep learning algorithm for automated design of druglike analogues[J]. J Chem Inf Model, 2022, 62(6): 1411-1424. doi: 10.1021/acs.jcim.2c00205
[58] Born J, Manica M. Regression Transformer enables concurrent sequence regression and generation for molecular language modelling[J]. Nat Mach Intell, 2023, 5(4): 432-444. doi: 10.1038/s42256-023-00639-z
[59] Wang JK, Zeng YD, Sun HY, et al. Molecular generation with reduced labeling through constraint architecture[J]. J Chem Inf Model, 2023, 63(11): 3319-3327. doi: 10.1021/acs.jcim.3c00579
[60] Du HY, Jiang DJ, Zhang O, et al. A flexible data-free framework for structure-based de novo drug design with reinforcement learning[J]. Chem Sci, 2023, 14(43): 12166-12181. doi: 10.1039/D3SC04091G
[61] Liao ZR, Xie L, Mamitsuka H, et al. Sc2Mol: a scaffold-based two-step molecule generator with variational autoencoder and transformer[J]. Bioinformatics, 2023, 39(1): btac814. doi: 10.1093/bioinformatics/btac814
[62] Liu XH, Ye K, van Vlijmen HWT, et al. DrugEx v3: scaffold-constrained drug design with graph transformer-based reinforcement learning[J]. J Cheminf, 2023, 15(1): 24. doi: 10.1186/s13321-023-00694-z
[63] Zhu HM, Zhou RY, Cao DS, et al. A pharmacophore-guided deep learning approach for bioactive molecular generation[J]. Nat Commun, 2023, 14(1): 6234. doi: 10.1038/s41467-023-41454-9
[64] Imrie F, Hadfield TE, Bradley AR, et al. Deep generative design with 3D pharmacophoric constraints[J]. Chem Sci, 2021, 12(43): 14577-14589. doi: 10.1039/D1SC02436A
计量
- 文章访问数: 514
- HTML全文浏览量: 130
- PDF下载量: 132
