1. 성봉준 교수 

Molecular simulation and artificial intelligence model studies of the packaging and the ejection processes of DNA in Viral Capsids

Abstract: 

DNAs are packaged into and ejected from a small viral capsid during the virus replication cycle. Single molecule experiments were performed to understand how the DNA could overcome a large entropy loss and a strong pressure during packaging. Theoretical studies also reported how the DNA was jammed in non-equilibrium states inside the capsid. It is also an issue how such non-equilibrium jammed conformation would affect the DNA ejection from a viral capsid. Unfortunately, however, packaging and ejection processes have been treated as independent processes under the assumption that the DNA would reach an equilibrium conformation. In this presentation, I would like to report that the ejection process of DNA from the viral capsid should depend significantly on how the DNA was packaged into the viral capsid. We perform Langevin dynamics simulation to package the DNA into a viral capsid and let the DNA eject from the capsid spontaneously. We find from our simulations that there should be three different regimes depending on the packaging rate: (1) knot dominant, (2) non-equilibrium dominant, and (3) intermediate regimes. When the DNA is packaged slowly, the DNA forms a complex knot easily during the packaging such that the ejection slows down (knot dominant regime). When the DNA is packaged quickly, the DNA is more likely to be jammed in non-equilibrium states, slowing down the ejection process (non-equilibrium dominant regime). When the packaging rate is intermediate, the probability of knot conformation is relatively low, and the DNA conformation may also relax easily, which facilitate the ejection most (intermediate regime). This indicates that the non-equilibrium conformation of DNA generated during the packaging (including knots) should affect the ejection kinetics significantly. We also report our recent convolutional neural network (CNN) model to identify and classify knots from the contact maps of DNA.

2. 김준수 교수

Pattern Formation on Phase-separating Lipid Vesicles by Uniaxial Compression

Abstract: 

The formation of patterned domains on particle surfaces is crucial for controlling particle self-assembly. In this work, we use molecular dynamics (MD) simulations to demonstrate that strong uniaxial compression of spontaneously phase-separating ternary lipid vesicles induces irreversible lipid-domain pattern formation. Using the MARTINI coarse-grained model, we simulated the spontaneous phase separation of a ternary lipid mixture (DIPC, DPPC, and cholesterol) into distinct domains on lipid vesicles. Uniaxial compression generates a discoid morphology with flat top and bottom planes and a curved edge, driving lipid-domain segregation between the planar and edge regions based on the strength of compression: moderate compression induces the band formation of DPPC/cholesterol at the edge, while strong compression replaces it with a band of DIPC at this region. Notably, the discoid vesicles with a DIPC band at the edge, induced by strong uniaxial compression, retain their shape and lipid-domain segregation even after the release of compressive forces, indicating irreversible pattern formation. Our findings suggest that external compression of small lipid vesicles provides a novel strategy for engineering surface patterns on nanoparticles. 

20250212_대학원특별세미나_성봉준교수_김준수교수.pdf