新型智算平台可精准预测与定向设计晶体机械性质

Core Insights - Tianjin University has developed an intelligent computing platform named CrystalGAT, which innovatively integrates graph attention neural networks with crystal engineering technology to accurately predict and design the mechanical properties of flexible crystal materials [1][2] - The platform significantly accelerates the traditional research process from "months to screen one effective structure" to "hundreds of candidate molecular libraries in a day," achieving a comprehensive accuracy rate of 90% in validation sets [1][2] Group 1: Technological Advancements - CrystalGAT achieves precise prediction and rational design of three mechanical properties: elasticity, plasticity, and brittleness of organic molecular crystals [2] - The platform constructs a full-chain technical paradigm of "data-driven - intelligent prediction - target identification - structure regulation," enabling researchers to lock in molecular modification targets and transition from blind exploration to precise optimization [2] Group 2: Practical Applications - In the pharmaceutical engineering field, the platform has identified two plastic co-crystals of the anti-epileptic drug Gabapentin, enhancing the tensile strength of tablets by 8.5 times and 5.7 times compared to the raw drug, effectively addressing the industry's challenge of fragility in tablet formation [2] - In functional materials, the team successfully transformed the brittle crystal PAPA into a flexible luminescent crystal that combines elasticity and light responsiveness, paving the way for new optical drive devices [2] Group 3: Accessibility and Collaboration - The platform is open-sourced on Hugging Face, allowing global researchers to access it without programming skills; they can simply draw or input molecular structures to obtain property predictions and visualizations of key segments [2]