Roadmap

Roadmap for Alchemlyb

Aims

1. To provide a modular, robust, and rigorously tested library for alchemical free energy calculations.
2. To provide state-of-the-art and best-practice implementations of algorithms for (1) pre-processing and (2) estimation
3. To provide modular tools for the analysis of sampling and convergence of alchemical free energy calculations.
4. To make it easy to couple structural analysis (e.g. conformations, dihedrals, other arbitrary coordinate-based features, solvation, ligand degrees of freedom) with alchemical analysis
5. To become a sustainable, community-driven project.

The foundation is Aim 1 but everyone on the 2017-10-30 call agreed that 2-4 are necessary to drive adoption and give people an immediate reason to use it. 5 is important for the long-term health of the project (and wide adoption).

Test-driven development is key (see Developer Guide) (to fulfill Aim 1) but also as a means to allow different contributors to add code to the library without fear of breaking working code, thus enabling Aim 5.

Interested parties

• Oliver Beckstein @orbeckst, Arizona State University
  • currently maintaining the library (in stewardship mode...)
  • uses it in-house
• Michael Shirts, University of Colorado Boulder
  • Working to get the nice features of alchemical-analysis on top of an extensible library.
• Bryce Allen, STX, @brycestx
  • Somewhat limited bandwidth but working with @shuail and @marcelschumann2 to contribute amber parsers and tests.
• David Mobley, UCI @davidlmobley
  • Limited bandwidth at present, but ultimately someone from the group will likely be involved, plus I can pass on lessons learned, things to avoid, etc.
  • Very involved with legacy alchemical analysis, but also currently afraid of changing anything in it because of the lack of continuous integration testing to ensure that changes don't break it (bad) or make it incorrect (worse).
• Shuai Liu, STX, @shuail
  • add the amber TI parser and test code for the parser
  • Long term: migrate other functions from alchemical-analysis
• Dominik Wille, Freie Universität Berlin @harlor
  • alchemical-analysis successor (currently developed as flamel)

Immediate/pressing requirements

parsers

• Amber
  • TI #10
  • FEP/MBAR #42
• Gromacs extended ensemble #14:
  • lambda exchange
• NAMD #7

(Note that each must be accompanied by tests; Ideally, tests should be written first to allow for test-driven development.)

docs

Howtos

• Show how it is already possible to couple conformational analysis to alechemical analysis. alchemlyb already makes it easy to pre-process "alchemical" dataframes (dH/dl or u_nk timeseries) by slicing but this needs to be shown-off (examples, tutorial).

Tutorials

• Show how to apply alchemlyb-based workflows to real-world examples (e.g., model systems that @brycestx mentioned).

test data

We need small data sets for test driven development. Data sets are packaged separately in the alchemistry/alchemtest package. See contributing data to alchemtest for further details.

Data needed:

• Amber TI #10
• Gromacs extended ensemble #14
• NAMD alchemtest #17

Librarify alchemical-analysis functionality

alchemical-analysis announced migration to alchemlyb. Prioritization of alchemical-analysis features to be migrated is an ongoing (early 2018) discussion.

As discussed in MobleyLab/alchemical-analysis#111, development of the alchemical-analysis successor will be based on flamel; once this is ready, it will be re-branded as alchemical-analysis.

Longer-term requirements

• Data for Chodera's Yank; coordinate with Chodera lab?