Validation of Occupancy Images from PET Data. A Novel Endpoint for Drug Discovery

摘要 PET成像在中枢神经系统药物开发中最有用的应用是测量新药的占有率 候选的靶结合部位(例如，受体、转运体和酶)。这些目标入住率(TO) 研究通常从灵长类动物开始，在人类身上进行到第一阶段-对于体内剂量非常有帮助- 寻找、决定是否将候选化合物推进到成本更高的后期试验，并优化 后期研究的设计。技术现状：一种常见的、信息量丰富的方法来分析 对数据来说就是“拉森阴谋”。Lassen图产生两点估计：目标的占有率 通过药物(ODrug)和示踪剂的不可置换摄取(VND)。这种流行背后的假设 方法是(A)这两种测量在整个大脑中是一致的，(B)PET示踪剂结合到 与候选药物相同的靶点人群。技术差距：有一些重要案例 违反了拉森情节的哪一个或多个假设。在这种情况下，(1)要研究的结果 可能有偏见，(2)可能需要人工干预，(3)正确的解释将取决于之前 目标亚型的空间分布知识，在任何情况下，(4)标准方法几乎不能提供 没有关于ODrug的地区差异的信息。 实验室的基本发展：在我们耶鲁大学的研究小组中，我们开发了一个扩展 标准拉森图，提供有关ODrug和VND的局部变异的信息。我们的结果是 新的Lassen Plot Filter(LPF)算法是ODrug、VND和EC50的逐个体素估计(即 图像“，”不可移位摄取，以及药物亲和力，图像“)。我们相信这些新奇的图像代表了许多 TO研究的结果比识别最大值的精确位置的标准测量结果更有信息量 药物的特定作用。例如，当假设被违反时，入住率图像将较少偏向 而不是点估计值。新的映像可以作为更丰富的端点来推动通过/不通过决策 候选化合物基于药物在特定大脑位置的作用，具有最大的治疗意义。 学术-产业合作伙伴关系。我们的工业合作伙伴，一家拥有丰富经验的公司 将为我们提供非常有价值的人类和非人类灵长类动物的档案数据，用于测试和 验证我们的LPF算法。 在本项目中，我们将(1)分析已知入住率分布的模拟数据。(2)重新分析 代表不同案例和/或违反标准假设的档案数据。(三)演出 在灵长类动物中进行的受限研究，以证实我们新图像的生物学解释。这项工作将是 当我们已经充分表征和优化了LPF算法的性能之后，我们就完成了。在漫长的岁月里 学期中，我们的目标是对TO研究的逐个体素分析的验证版本，并广泛采用，以 最终用户的利益，以加快药物开发。

Abstract The most useful application of PET imaging in CNS drug development is to measure occupancy of new drug candidates at target binding sites (e.g., receptors, transporters, and enzymes). These target occupancy (TO) studies -often starting in primates and progressing to Phase 1 in humans - are very helpful for in vivo dose- finding, deciding whether to advance a candidate compound to more costly later phase trials, and optimizing the design of later phase studies. Current State of the Art: A common and informative approach to analysis of TO data is the “Lassen plot”. The Lassen plot yields two point estimates: a fractional occupancy of the target by the drug (ODrug) and the nondisplaceable uptake of the tracer (VND). Assumptions underlying this popular method are that (a) both measures are uniform across the whole brain and (b) the PET tracer binds to the identical population of target sites as the candidate drug. Technology Gap: There are important cases in which one or more assumption of the Lassen plot is violated. In such cases, (1) the outcome of the TO study may be biased, (2) manual intervention may be required, (3) proper interpretation will depend on prior knowledge of the spatial distribution of target subtypes, and in any case, (4) the standard method offers little to no information on regional variation in ODrug. Basic development in the lab: In our research group at Yale, we have developed an extension of the standard Lassen plot that provides information about local variation in ODrug and VND. The outcomes of our new “Lassen Plot Filter” (LPF) algorithm are voxel-by-voxel estimates of ODrug, VND and EC50 (i.e., `occupancy images', `nondisplaceable uptake, and drug affinity, images'). We believe these novel images represent much more informative outcomes of TO studies than standard measures for identifying precise locations of maximal specific action of drugs. For example, when assumptions are violated, occupancy images will be less biased than point estimates. The new images could serve as richer endpoints to drive go/no-go decisions on candidate compounds based on drug action at specific brain locations of greatest therapeutic interest. Academic-Industrial Partnership. Our industrial partner, a company with extensive experience in conducting TO studies, will provide us with very valuable archival data in humans and nonhuman primates for testing and validating our LPF algorithm. In the present project, we will (1) Analyze simulated data with known occupancy distributions. (2) Re-analyze archival data that represent different cases and/or violations of standard assumptions. (3) Perform circumscribed studies in primates to confirm the biological interpretations of our new images. The work will be complete when we have fully characterized and optimized the performance of our LPF algorithm. In the long term, our goal is for a validated version of our voxel-by-voxel analysis of TO studies, to be adopted widely, to the benefit of end-users, to speed drug development.