[fizixer]

It's two separate things, both mentioned in your question: simulation then animation. Both steps are computationally expensive enough that they can't be done in one run.

For simulation, typically a molecular dynamics code is used along with features specific to biological analysis (since MD by itself is a physics simulator). NAMD is one of the well-known MD tools used for bio work. Depending on the complexity and features/accuracy sought, a whole host of other tools might be involved. The timeframe of simulation is of the order of weeks to months given massive computational power (e.g., a large allocation on one of the national supercomputing grids).

Once tons of simulation data is generated (as the output of simulation), then 3D visualization tools can be used. NAMD has a sister vis tool called VMD. Blender is another option. Incidentally, Janet Iwasa (the narrator and researcher behind this work) is a visualization expert, so it's likely she only worked on the second step (animation) by using existing simulation data, or collaborating with a simulation group. (Again the computation required for rendering is very high, on the order of days to weeks using a large allocation on a supercomputer).