**佐治亚理工联合多伦多大学与康奈尔大学推出全新分子优化技术MOLLEO**
近日，佐治亚理工学院、多伦多大学和康奈尔大学的研究者们宣布，他们合作开发出一种名为MOLLEO的分子语言增强进化优化技术。这一技术显著改善了进化算法在分子发现领域的优化能力，被认为是化学领域的一大突破。
分子发现过程一直面临巨大的计算挑战，由于涉及的复杂设计条件和分子属性的评估通常需要昂贵的实验和模拟。佐治亚理工学院的王浩瑞及其团队意识到这一问题后，开始探索新的解决方案。他们决定将预训练的大语言模型整合到进化算法中，旨在通过自然语言处理技术加速分子优化过程。
此次推出的MOLLEO技术便是这一思路的结晶。该技术的核心在于利用拥有化学知识的大语言模型来指导进化算法的搜索过程，使进化算法能够更有效地遍历化学空间，减少了昂贵目标评估的需求。这一技术有望大幅加快药物设计、材料研发等领域的进程。
该研究成果以《Efficient Evolutionary Search Over Chemical Space with Large Language Models》为题，于近日发表在预印平台arXiv上，论文链接已附在文中。专家表示，这一技术可能是解决分子发现计算挑战的关键一步，并可能推动化学领域的快速发展。
英语如下：
News Title: “MOLLEO Language Model Developed by Georgia Tech and Other Universities Revolutionizes Molecular Discovery and Optimization Processes”
Keywords: MOLLEO, Evolutionary Algorithm Improvement, Large Language Model Integration
News Content: **Georgia Tech Joins Forces with University of Toronto and Cornell to Introduce New Molecular Optimization Technology, MOLLEO**
Recently, researchers from Georgia Institute of Technology, the University of Toronto, and Cornell University announced their collaboration to develop a new technology called MOLLEO, which stands for Molecular Language Enhanced Evolutionary Optimization. This technology significantly improves the optimization capabilities of evolutionary algorithms in the field of molecular discovery and is considered a breakthrough in the chemical industry.
The process of molecular discovery has always faced tremendous computational challenges, as complex design conditions and evaluations of molecular properties often require expensive experiments and simulations. After realizing this issue, Wang Haorui and his team at Georgia Tech began exploring new solutions. They decided to integrate pre-trained large language models into evolutionary algorithms, aiming to accelerate the process of molecular optimization through natural language processing techniques.
The newly introduced MOLLEO technology is the crystallization of this approach. At the core of this technology is the utilization of large language models with chemical knowledge to guide the search process of evolutionary algorithms, enabling them to traverse the chemical space more efficiently and reducing the need for expensive target evaluations. This technology has the potential to significantly accelerate processes in drug design, material development, and other fields.
The research findings were published on the preprint platform arXiv under the title “Efficient Evolutionary Search Over Chemical Space with Large Language Models.” Links to the paper are provided in the article. Experts indicate that this technology could be a crucial step in addressing the computational challenges of molecular discovery and may propel the rapid development of the chemical industry.
【来源】https://www.jiqizhixin.com/articles/2024-07-01-15
Views: 1