Traditionally free energy of biomolecular systems is divided into gas phase contribution and solvation free energy, which is further split into two independent contributions (polar and non-polar). Such formulation increases variance of calculation on the one hand, and has no direct experimental validation on the other hand. The generalized solvation free energy (GSFE) is proposed to overcome these limitations. In GSFE, each comprising unit is both a solute and part of the solvent for its neighboring unit. Consequently, solvent is generally heterogeneous and specific for each unit. The definition of comprising unit is inherently flexible and multi-scale (e.g. atoms, atomic groups, amino acids and/or their clusters etc.). GSFE is first implemented for proteins at residue level with neural network and utilized for assessment of protein structural models, structural refinement and design. Implementation of GSFE at atomistic level is undergoing.

2003年获得犹他大学材料科学与工程专业博士学位，2003-2005，密歇根大学博士后；2005-2009，美国国立卫生研究院博士后；2009-今，吉林大学生命科学学院唐敖庆特聘教授。研究内容包括：在蛋白质自由能计算和动态关联分析方面开展了原始理论创新研究。在过去的三年中基本完成了由传统统计热力学框架向机器学习（人工智能）框架的转变。目前致力于开发下一代融合人工智能与统计热力学理论，能够在很大程度上取代目前众多定性和部分定量实验的自学习生物大分子相互作用预测平台，从而解除目前生物医药研究主要瓶颈，即分子相互作用的准确预测。