SBIR Phase I: A physics-based machine learning platform for crystal structure prediction of small drug molecules

这个小企业创新研究（SBIR）第一阶段项目的广泛影响是加速和降低小分子药物研究早期阶段的成本。制药公司能够推向市场的药物数量受到开发每种药物所涉及的时间、成本和复杂性的限制。研发过程通常需要10年左右的时间，每年很少有药物能进入市场。这项技术在提高小分子药物的开发频率方面可能特别有影响力，这些药物用于治疗未充分研究的疾病，这些疾病总共影响了3000多万美国人。通过降低新药物的成本和上市时间，该项目可以推动该行业的发展，并为那些目前没有药物选择的疾病患者带来改变生活的治疗方法。本项目开发解决晶体结构预测（CSP）问题的技术。小分子和多肽的晶体结构决定了许多药理学特性，包括溶解度、口服生物利用度、保质期稳定性和毒性。晶体结构的实验测定是昂贵的，需要大量的人力来进行，因此计算方法将重塑小分子药物的特征。提出的技术创新结合了一种基于量子化学的新型能量预测模型和一种机器学习方法，可以有效地对可能的晶体结构的广阔空间进行采样。由此产生的技术将帮助制药公司降低药物开发过程中的风险，因为它允许制药公司在实验室合成药物之前对晶体结构进行计算分析。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 accelerate and reduce the cost of the early stages of small molecule pharmaceutical research. The number of drugs a pharmaceutical company can bring to market is limited by the time, cost, and complexity involved in developing each drug. The research and development process typically takes around 10 years, and few drugs make it onto the market each year. This technology may be especially impactful in improving the frequency at which small molecule drugs are developed for understudied diseases, which collectively impact over 30 million Americans. By reducing the cost and time to market for new pharmaceuticals, the project could advance the industry and bring life-changing therapeutics to underserved people who are suffering from illnesses where there are presently no drug options.This project develops technologies to solve the crystal structure prediction (CSP) problem. The crystalline structure of small molecules and peptides determines many pharmacological characteristics including solubility, oral bioavailability, shelf-life stability, and toxicity. Experimental determination of the crystal structure is expensive and requires significant human labor to conduct, so a computational approach would reinvent the characterization of small molecule drugs. The proposed technical innovation combines a novel energy prediction models based on quantum chemistry with a machine learning method for efficiently sampling the vast space of possible crystal structures. The resulting technology will help pharmaceutical companies de-risk their drug development process by allowing them to analyze crystal structures computationally before having to synthesize them in the lab.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.