Learned replica exchange (LREX)

Code to reproduce the paper Skipping the replica exchange ladder with normalizing flows, JPCL (2022). [Supporting Information]

The systems studied are: a multidimensional double-well (dw), alanine dipeptide (ala2), and alanine tetrapeptide (ala4).

Requirements

• bgflow
• bgmol
• openmm
• plumed and openmm-plumed for the reference simulations with OPES

Quick start

A self-contained Google Colab notebook is also available:

Errata corrige

There is an error in the paragraph just after eq 3 of the paper. For two times the words "prior" and "target" have been switched.

Thus $x_p \sim p(x)$ and $x_q \sim q(x)$, and the target continues from $f(x_q)$ and the prior from $f^{-1}(x_p)$, not vice versa as wrongly stated in the paper.