SBIR Phase I: Developing NLP-based Platform Technology for Phage-Host Interaction Predictions

这项小企业创新研究（SBIR）第一阶段项目的更广泛影响是改善与感染有关的农业经营。农作物对抗生素的耐药性正在增加。最有希望的抗生素替代品之一是噬菌体，也被称为噬菌体。噬菌体是一种安全的细菌病毒，已经在美国进行了小规模的商业应用。在大量疾病中使用噬菌体的一个主要瓶颈是难以选择对目标病原体有效的噬菌体。该项目旨在开发一种最先进的计算方法，以快速有效的方式选择噬菌体对抗一系列疾病，从而解决了这一挑战。这项技术将使噬菌体作为一种更安全、更有效的抗生素替代品得到广泛应用。此外，该项目的最终成果还包括开发出有史以来第一个基于噬菌体的、可持续的、具有成本效益的抗黄龙病抗菌剂，黄龙病是一种毁灭性的、完全无法治愈的柑橘细菌性疾病。拟议的项目旨在通过建立第一个商业上可行的平台技术来计算预测噬菌体-宿主相互作用，从而消除噬菌体领域的一个关键瓶颈。具体来说，目前选择针对目标病原体的噬菌体是困难和耗时的，因为这种人工过程效率低下。此外，噬菌体不能从不可培养的细菌中分离出来。拟议的项目旨在利用尖端的自然语言处理技术，用于噬菌体对抗越来越多的病原体。这项提议的第一个目标是显著改进已经建立的模型，并表明他们的预测在体外使用金标准斑块测定实验是成立的。该提案的第二个目标是使用该平台技术和湿实验室方法生产和测试有史以来第一个抗菌（基于噬菌体的）溶液，该溶液可用于治疗由不可培养病原体引起的目前无法治愈的黄龙冰病。该项目将进一步帮助确定基于transformer的模型是否能够在复杂的基因组理解任务中表现出色，并产生体外可行的病毒-宿主相互作用预测。该奖项反映了美国国家科学基金会的法定使命，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 of this Small Business Innovation Research (SBIR) Phase I project is to improve agricultural operations with respect to infection. Antibiotic resistance in crops is increasing. One of the most promising alternatives to antibiotics is bacteriophages, also known as phages. Phages are safe bacterial viruses already used on a small commercial scale in the United States. One major bottleneck for phage use in a larger number of diseases is the difficulty involved in selecting phages that will be effective against the target pathogen. This project addresses this challenge by aiming to develop a state-of-the-art computational approach to select phages against a range of diseases in a rapid and efficient manner. This technology will allow for widespread phage use as a safer and more efficient substitution for antibiotics. Additionally, the final deliverable of this project includes developing the first-ever phage-based, sustainable, and cost-effective antimicrobial solution against Huanglongbing, a devastating and completely incurable bacterial disease of citruses.The proposed project aims to remove a key bottleneck in the bacteriophage field by building the first-ever commercially viable platform technology to computationally predict phage-host interactions. Specifically, it is currently difficult and time-consuming to select bacteriophages against target pathogens as this manual process is inefficient. Moreover, phages cannot be isolated against uncultivable bacteria. The proposed project aims to use cutting-edge Natural Language Processing for bacteriophage use against an increasing number of pathogens. This proposal’s first objective is to significantly improve the already built models and show that their predictions hold in vitro using the gold-standard plaque assay experiments. This proposal’s second objective is to use this platform technology and wet-lab methods to produce and test the first-ever antibacterial (phage-based) solution that works against the currently incurable Huanglongbing disease caused by an uncultivable pathogen. This project will further help determine if Transformer-based models can excel at complex genome understanding tasks and produce in-vitro-viable viral-host interaction predictions.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.