Adversarial reverse mapping of condensed-phase molecular structures: Chemical transferability

Switching between different levels of resolution is essential for multiscale modeling, but restoring details at higher remains challenging. In our previous study, we have introduced deepBackmap, a deep neural-network-based approach to reverse-map equilibrated molecular structures condensed-phase systems. Our method combines data-driven and physics-based aspects, leading high-quality reconstructed structures. this work, expand the scope model examine its chemical transferability. To end, train deepBackmap solely on homogeneous liquids small molecules apply it more challenging polymer melt. We augment generator’s objective with force-field-based terms as prior regularize results. The best performing physical depends whether specific chemistry or transfer model. local environment representation combined sequential reconstruction fine-grained helps in reaching transferability learned correlations.