Modular reweighting software for statistical mechanical analysis of biased equilibrium data

Abstract

Here a simple, useful, modular approach and software suite designed for statistical reweighting and analysis of equilibrium ensembles is presented. Statistical reweighting is useful and sometimes necessary for analysis of equilibrium enhanced sampling methods, such as umbrella sampling or replica exchange, and also in experimental cases where biasing factors are explicitly known. Essentially, statistical reweighting allows extrapolation of data from one or more equilibrium ensembles to another. Here, the fundamental separable steps of statistical reweighting are broken up into modules – allowing for application to the general case and avoiding the black-box nature of some “all-inclusive” reweighting programs. Additionally, the programs included are, by-design, written with little dependencies. The compilers required are either pre-installed on most systems, or freely available for download with minimal trouble. Examples of the use of this suite applied to umbrella sampling and replica exchange molecular dynamics simulations will be shown along with advice on how to apply it in the general case. New version program summaryProgram title: Modular reweighting version 2Catalogue identifier: AEJH_v2_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEJH_v2_0.htmlProgram obtainable from: CPC Program Library, Queenʼs University, Belfast, N. IrelandLicensing provisions: GNU General Public License, version 3No. of lines in distributed program, including test data, etc.: 179 118No. of bytes in distributed program, including test data, etc.: 8 518 178Distribution format: tar.gzProgramming language: C++, Python 2.6+, Perl 5+Computer: AnyOperating system: AnyRAM: 50–500 MBSupplementary material: An updated version of the original manuscript (Comput. Phys. Commun. 182 (2011) 2227) is availableClassification: 4.13Catalogue identifier of previous version: AEJH_v1_0Journal reference of previous version: Comput. Phys. Commun. 182 (2011) 2227Does the new version supersede the previous version?: YesNature of problem: While equilibrium reweighting is ubiquitous, there are no public programs available to perform the reweighting in the general case. Further, specific programs often suffer from many library dependencies and numerical instability.Solution method: This package is written in a modular format that allows for easy applicability of reweighting in the general case. Modules are small, numerically stable, and require minimal libraries.Reasons for new version: Some minor bugs, some upgrades needed, error analysis added.Summary of revisions:1.analyzeweight.py/analyzeweight.py2 has been replaced by “multihist.py”. This new program performs all the functions of its predecessor while being versatile enough to handle other types of histograms and probability analysis.2.“bootstrap.py” was added. This script performs basic bootstrap resampling allowing for error analysis of data.3.“avg_dev_distribution.py” was added. This program computes the averages and standard deviations of multiple distributions, making error analysis (e.g. from bootstrap resampling) easier to visualize.4.WRE.cpp was slightly modified purely for cosmetic reasons.5.The manual was updated for clarity and to reflect version updates. Examples were removed from the manual in favor of online tutorials (packaged examples remain).6.Examples were updated to reflect the new format. An additional example is included to demonstrate error analysis.Running time: Preprocessing scripts 1–5 minutes, WHAM engine <1 minute, postprocess script ∼1–5 minutes.