I-Corps: Application of deep generative models for simulating biological systems

这个I-Corps项目更广泛的影响/商业潜力是为模拟生物系统开发基于人工智能（AI）的平台，该平台可以在湿实验室实验之前预测任何扰动的真实表型结果。所提出的技术的开发解决了对高度预测性、高效且具有成本效益的平台的需求，该平台在生物制药行业的药物发现、基因治疗和个性化医疗中具有潜在的应用。传统上，药物、疫苗和疗法的开发是在生物湿实验室环境中进行的，从临床前细胞和动物模型到临床阶段的人体试验。这项研究往往是昂贵的，需要多年的努力，并可能难以达到适当的疗效。该项目利用人工智能（AI）和机器学习（ML）技术，通过深度生成模型（DGMs）加速生物系统中表型结果的预测。深度神经网络与随机优化方法的进步相结合，实现了复杂高维数据的可扩展建模，并已成为计算机视觉、语音和自然语言处理、图形挖掘和强化学习中通常首选的人工智能方法。 然而，很少有应用于生物系统的例子。该项目结合了生物过程和DGM来训练深度神经网络，以表征每个过程特有的表型。然后，经过训练的DGMs可以预测在任何给定环境中对这些过程的任何扰动的最佳结果。该技术具有可重复性和可扩展性，旨在提供生物和药理物质作用的高含量结构、表型和形态学特征。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of an artificial intelligence (AI) based platform for simulated biological systems that can predict the true phenotypic outcome of any perturbation prior to wet lab experimentation. The development of the proposed technology addresses the need for a highly predictive, efficient, and cost-effective platform that has potential applications in drug discovery, gene therapy and personalized medicine in the biopharmaceutical industry. Traditionally, the development of drugs, vaccines, and therapies is carried out in biological wet lab settings from preclinical cell and animal models to clinical phase human trials. This research is often expensive, requires years of effort, and can struggle to achieve suitable efficacy. The proposed technology may offer prediction of biochemical changes with high accuracy and maximum efficacy and safety, thereby reducing the burden on payers and stakeholders.This I-Corps project leverages artificial intelligence (AI) and machine learning (ML) through deep generative models (DGMs) to accelerate prediction of phenotypic outcomes in biological systems. Deep neural networks combined with progress in stochastic optimization methods have enabled scalable modeling of complex, high-dimensional data and has become the often preferred artificial intelligence method in computer vision, speech and natural language processing, graph mining, and reinforcement learning. However, there are few examples of its application to biological systems. This project combines the biological processes and DGMs to train the deep neural network in characterizing the phenotype unique to each process. The trained DGMs can then predict the best outcome for any perturbation to these processes in any given environment. The proposed technology is reproducible and scalable, and designed to provide high-content structural, phenotypic, and morphological profiles of the effects of biological and pharmacological substances.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.