基于人工智能与大脑活动图谱的新型高通量药物筛选研究

Millions of people in China are suffering from the brain diseases. Despite the increasing R&D investment of the corresponding drugs for brain diseases, few new drugs have been discovered. It is resulted from the complexity of the human brain and the various technical bottlenecks in drug development. Technologies for mapping the patterns of neural activity have advanced our understanding of brain function in disease. An important application of these technologies is the discovery of new drugs or neurotherapeutics for brain disorders. Many studies have focused on the disease treatment based on brain activity imaging (BAMing) technology, but few has been devoted to drug discovery. There are several technical challenges, including low throughput, feature extraction of multidimensional maps and the functional prediction mechanisms, which need to be addressed before we can apply BAMing technology in drug discovery. Here, a high-throughput drug screening platform is proposed based on artificial intelligence and BAMing. This system combines microfluidic chips and neuro-imaging technique to enable high-throughput BAMing. Lage scale of multidimensional brain activity maps (BAMs) will be generated, to form a database of information-rich BAMs. Based on unsupervised machine learning, a precise and efficient drug prediction mechanism will be established to identify the new functional drugs for brain diseases. Their therapeutic effects will also be validated in animal disease models in the final. Collectively, this study will provide a framework to advance the field of next-generation drug discovery.

大脑疾病的治疗正困扰着数以百万计的中国人民。虽然对相应药物的研发投入逐年增加，但新药发现的速度却持续放缓。这归咎于人类大脑的复杂以及药物研发技术的各式瓶颈。随着神经活动图谱绘制技术的进步，病患的大脑功能得到较深层次的认知。这些技术的一个重要应用便是帮助我们找寻治疗大脑疾病的新药或新方法。前人对利用大脑活动成像技术进行疾病治疗投入了大量的研究工作，但用以药物研发却鲜有报道。其中有多个关键技术难题亟需解决，包括技术高通量化，多维图谱的特征提取，预测机制的高效精准化等。本研究计划基于人工智能与大脑活动图谱绘制技术，结合微流控芯片和神经成像，开发高通量的新型药物筛选平台；进行大规模药物测试，相应地绘制多维大脑活动图谱，形成富含药物评价信息的数据库；通过机器学习分析方法，建立精准高效的药物预测机制，同时筛选出新型治疗药物，并将于动物疾病模型上进行验证，为新一代药物研发提供强有力的研究依据。

具备时空分辨率的神经活动图谱绘制技术促进了我们对健康和疾病状态下大脑功能的理解。 这些技术的一个重要应用是用于发现神经和精神疾病的新一代治疗方法。在本项目的资助下，我们描述了一种在体的药物筛选策略。该策略结合了用于生成大规模大脑活动图谱的高通量筛选技术和用于预测分析的机器学习等人工智能算法。具体而言，本研究描述了一种新型筛选系统，它结合了自动斑马鱼封装技术和液滴微阵列策略，用于单药/多药的体内功能筛选。该平台可通过使用微流体装置在琼脂糖中实现自动化、快速的斑马鱼定向和固定，以生成大批量的斑马鱼胶囊。结合机器学习算法，该平台能够根据药物处理后所生成的内涵信息丰富的斑马鱼幼鱼大脑活动图谱，评估化合物的作用机制和潜在治疗用途。从已临床使用的药物筛选中，我们发现了与已知治疗类别存在内在联系的药物聚类。利用基于大脑活动图谱的功能分类器，我们初步筛选出抗癫痫药物先导物，及能够诱导出与单药相似的神经调节效果但无明显心脏毒副作用的药物组合。相信该策略可以很容易地应用于多个领域，特别适用于资源与样品有限的单一/组合药物的探索，以加速精准药物新疗法的发现。