Motion Sequencing for Neuropsychiatric Drug Development

总结 神经精神疾病折磨着全球超过20%的人口，导致巨大的个人和社会问题。 社会负担，包括数万亿美元的总成本。尽管其流行和广泛的影响， 治疗神经精神疾病的药物开发明显落后于其他疾病领域。这一差距 主要是因为很少有候选药物进入临床;这些药物的平均滞后时间 这确实使它成为13年，进一步加剧了问题。精神科药物的成功率在历史上 低，即使在该地区的金融投资上升。开发安全有效的神经精神药物 疾病的研究本质上是困难的，因为它依赖于表征动物模型的行为表型。电流 对此的方法是低通量、不可靠、昂贵和信息量最小的。大多数方法试图 将依赖于许多神经回路的复杂行为简化为一个或几个可量化的指标， 然后用来预测候选人将如何影响更复杂的人类神经系统。音节 生命科学是建立在改善我们测量和解释行为变化的方式的愿景之上的。 来改善临床前药物的开发。为了应对这一挑战，我们开发了一种行为 运动测序（MoSeq）。MoSeq结合了机器视觉和无监督 机器学习技术来客观地识别一组刻板的三维行为图案 （后仰、转身、甩头、跑动、停顿等）将老鼠所有的自发行为都包含在一个 特殊实验除了揭示哪一个主题（称为“行为音节”）在每个 目前，MoSeq确定了控制音节如何随时间从一个过渡到另一个的统计数据 （“行为语法”）。因此，使用MoSeq，我们可以全面定量地描述啮齿动物 行为我们之前的工作表明，MoSeq直接反映了精神病学中正在进行的大脑活动- 相关的大脑回路，并且它可能显着优于更标准的药物表型分析方法。 对小鼠的影响在这个SBIR第一阶段项目中，我们建议扩展MoSeq的功能，并明确 展示其翻译价值。具体来说，在目标1中，我们将扩大MoSeq的范围， 神经精神病学相关的电路;在目标2中，我们将建立一个行为空间，描述 在目标3中，我们将证明MoSeq的临床实用性 来预测临床试验的结果。该项目将为彻底改变 神经和精神治疗的临床前管道。

SUMMARY Neuropsychiatric disorders afflict more than 20% of the global population, resulting enormous personal and societal burdens, including trillions of dollars in total costs. Despite its prevalence and widespread impact, the development of drugs to treat neuropsychiatric diseases significantly lags behind other disease areas. This gap is due primarily to the fact that so few candidate drugs ever make it to the clinic; the average lag time for those that do make it is 13 years, further exacerbating the problem. The success rate for psychiatric drugs is historically low, even as financial investments in the area rise. Developing safe and effective drugs for neuropsychiatric disorders is inherently difficult, as it relies on characterizing the behavioral phenotypes of animal models. Current approaches to this are low-throughput, unreliable, expensive, and minimally informative. Most methods attempt to reduce complex behaviors that depend upon many neural circuits into one or a few quantifiable metrics, which are then used to predict how the candidates will impact the even more complex human nervous system. Syllable Life Sciences was founded on the vision of improving the way we measure and interpret changes in the behavior in the lab to improve pre-clinical drug development. To address this challenge, we have developed a behavioral analysis platform called Motion Sequencing (MoSeq). MoSeq combines machine vision and unsupervised machine learning techniques to objectively identify a set of stereotyped three-dimensional behavioral motifs (rears, turns, head-bobs, runs, pauses, etc.) that encapsulates all the spontaneous actions of mice within a particular experiment. In addition to revealing which motif (termed a “behavioral syllable”) is expressed at each moment, MoSeq identifies the statistics that govern how syllables transition from one to anther over time (“behavioral grammar”). Using MoSeq, therefore, we can comprehensively and quantitatively profile rodent behavior. Our previous work demonstrates that MoSeq directly reflects ongoing brain activity in psychiatry- relevant brain circuits, and that it may significantly outperform more standard methods of phenotyping drug effects in mice. In this SBIR Phase I project, we propose to extend the capabilities of MoSeq and explicitly demonstrate its translational value. Specifically, in Aim 1 we will expand the purview of MoSeq to include neuropsychiatry-relevant circuits; in Aim 2 we will build a behavioral space that describes relationships among drugs spanning the current psychopharmacopeia; and in Aim 3 we will demonstrate the clinical utility of MoSeq by using it to predict clinical trial outcomes. This project will lay essential groundwork for revolutionizing the preclinical pipeline for neuro- and psychotherapeutics.