Soft matters including a wide range of materials from liquids, polymers, gels, to biomolecules can be found in our daily life as well as advanced high-tech areas. Designing new soft materials is fundamental to the development of novel technologies for chemical and biological applications. Achieving targeted structures/properties of materials involves lots of experimental efforts, considering the large diversity of chemical functional groups and the complicated experimental conditions such as temperatures, pressures and electric fields, etc. Computational simulations are good and cheap tools to help understand the âfunctional groupsâ- structure-property relationships of materials and provide guidelines for new material design. Coarse-grained (CG) molecular dynamics (MD) simulations have been widely used to study soft material such as polymers and biomolecules. These macromolecules usually undergo complicated processes proceeding for tens to hundreds of microseconds. We firstly developed accurate coarse-grained models of a wide range of materials from water, hydrocarbons to polymers, i.e., poly(acrylic acid) (PAA) and polystyrene (PS). The newly developed polymer models were used to build the block polymers and bottlebrush copolymers. The self-assembly of block polymers and the conformation transition of bottlebrush copolymers were investigated by CGMD simulations. It was found the self-assembled nanospheres were obtained by the CG PAA-PS block models. The conformations of bottlebrush copolymers of PAA-PS were affected by the compositions of binary solvents of water and DMF, as well as the architectures of the bottlebrush copolymers. Although the CG models were useful in studying these macromolecules, the atomistic details were lost in these models. To gain the atomistic information, we developed machine learning algorithms to constructure all-atom models by backmapping their CG models. The reconstructed all-atom models were more accurate than those by using the algorithm of randomly fragment placement. My current research interests are CG modelling of biocondensates which can play pivotal roles in the formations of membraneless organelles. We aim to unveil the roles of amino acid sequences in determining the properties of the biocondensates. Overall, these computational studies could help understand the chemical and biological phenomena at microscopic level and further provide guidelines for material design when integrated with machine learning algorithms.