I work on methods for quantifying (and hopefully reducing!) sources of predictive uncertainty in molecular simulation. My main projects revolve around:
- measuring the sampling bias introduced by the possible procedures for discretizing the equations of motion, so we can identify procedures that introduce minimal bias,
- quantifying parameter uncertainty in molecular models, so we can propagate that uncertainty into predictions, and
- statistically compressing the long-timescale behavior of biomolecules from molecular dynamics trajectories, so we can make efficient use of finite simulation resources.