Fragment-to-lead and target validation

抗病毒疗法的发现需要一个对突变很强的靶点，以及合适的化学物质 调节目标的物质。片段到线索和目标验证项目处于至关重要的位置 目标验证和化学铅生成之间的接口。我们的目标是验证生物学假说 目标选择的背后，并为下游治疗开发生产铅分子，生产铅 对抗9个抗病毒目标。通过紧密集成超高通量X射线的独特能力 钻石光源的结晶学，这个项目利用了人工智能和 机器学习(Al/ML)和亿级计算自由能计算快速生成新的势能 能够克服来自初始X射线碎片筛选的阻力的先导化合物。该项目建立在 成功的COVID月球计划，执行了针对以下目标的快速碎片到领导运动 SARS-CoV-2主要蛋白酶-从片段筛选开始，得到一种先导化合物，其IC50=140 nM 在&lt；6个月内发现并制造了&lt；400个化合物。在Hit-to-Lead活动的第一阶段，我们将使用 机器学习从高通量片段屏幕读数中学习药效团特征，并使用 这些模式从虚拟的、可合成的化学空间中搜索有效的命中。我们的目标是 到达使病毒蛋白具有抗病毒活性的化学物质，这反过来又能 验证目标的实验。与项目1(抗病毒靶向抑制耐药性)合作， 有效的命中将被用来验证目标交战的生物学假设，以及深度突变扫描 并将使用连续传代来评估抗药性障碍和导致 抵抗。这些见解将被用于迭代药物化学设计过程，在选择 抗性潜力较小的化学系列，专注于扩展为突变靶标的载体 坚固的残留物。在打到领先运动的第二阶段，我们将在结构性和可持续发展的基础上 在第一阶段产生的生物活性数据，并使用机器学习、炼金术自由能计算，以及 高通量纳米分子化学，快速评估和合成类似物，扩展到 前景看好的载体。这一阶段的目标是：(I)IC50&lt；500生化分析中的有效抑制物 (Ii)EC503μM和细胞毒性CC50G50μM在细胞抗病毒检测中的抑制；和(Iii)可发展 一级ADME和理化性质：Clog P&lt；4，动力学溶解度&50μM，大鼠和人微粒体 稳定性：50、渗透率：1x108 cm/S。这些导线是导线优化的输入 (项目5)，该项目将侧重于进一步改进效力、admet和体内药代动力学。

The discovery of an antiviral therapeutic requires a target that is robust to mutations, and suitable chemical matter that modulates the target. The Fragment-to-Lead and Target Validation project sits at the crucial interface between target validation and chemical lead generation. We aim to validate biological hypotheses behind target selection and produce lead molecules for downstream therapeutic development, producing leads against 9 antiviral targets. By tightly integrating the unique capabilities of extremely high-throughput X-ray crystallography at Diamond Light Source, this project leverages recent advances in artificial intelligence and machine learning (Al/ML) and exascale computing free energy calculations to rapidly generate novel potent lead compounds able to overcome resistance from initial X-ray fragment screens. This project builds on the successful COVID Moonshot initiative, which executed a rapid fragment-to-lead campaign against SARS-CoV-2 main protease - starting from fragment screen, a lead compound with IC50 = 140 nM was discovered in <6 months and <400 compounds made. In the first stage of a hit-to-lead campaign, we will use machine learning to learn pharmacophore features from high throughput fragment screen readout, and use these patterns to search for potent hits from virtual, synthetically accessible chemical space. The goal is to arrive at chemical matter which engages the viral protein with antiviral activity, which in turn enables experiments that validate the target. Working with Project 1 (Antiviral Targeting to Suppress Resistance), potent hits will be used to validate biological hypotheses of target engagement, and Deep Mutational Scanning and serial passaging will be used to evaluate the barrier to resistance and the mutations that give rise to resistance. These insights will be used in the iterative medicinal chemistry design process, in selecting chemical series with less resistance potential and focusing on expanding into vectors that target mutationally robust residues. In the second stage of the hit-to-lead campaign, we will build on the wealth of structural and bioactivity data generated in the first stage and use machine learning, alchemical free energy calculations, and high throughput nanomole chemistry to rapidly evaluate and synthesize analogues which expand into promising vectors. The goal of this phase is to arrive at: (i) potent inhibitor in biochemical assays with IC50<500 nM; (ii) inhibition in cellular antiviral assays with EC50<3μM and cytotoxicity CC50>50 μM; and (iii) developable Tier 1 ADME and physicochemical properties: clog P<4, kinetic solubility > 50 μM, rat and human microsomal stability Clint<50, MDCK-LE permeability Papp>1x108cm/s. These leads are inputs to Lead Optimization (Project 5), which will focus on further improvements in potency, ADMET and in vivo pharmacokinetics.