Advancing Reproducibility and Open Data in Theoretical and Computational Chemistry
Rommie E. Amaro, Victor Batista, Jochen Blumberger, Yee Siew Choong, Clémence Corminboeuf, Zoe Cournia, Qiang Cui, Marco De Vivo, Francesco A. Evangelista, Yi Qin Gao, Debashree Ghosh, Xiao He, Olexandr Isayev, Syma Khalid, Johannes Kirchmair, John R. Kitchin, Haiyan Liu, Kevin J. Naidoo, Duc Nguyen, Ariane Nunes Alves, Giulia Palermo, Brett Savoie, Thereza A. Soares, Pratyush Tiwary, Guowei Wei, Xiao Zheng, Tong Zhu, Kenneth M. Merz, Jr. and Laura Gagliardi, “Advancing Reproducibility and Open Data in Theoretical and Computational Chemistry”, J. Chem. Theory Comput. 22(9), 4199-4200 (2026). https://doi.org/10.1021/acs.jctc.6c00733
This summary was written by a language model from the paper and has not been checked line by line. Rely on the published paper for details.
Question
This is a short joint editorial from the editors of the Journal of Chemical Theory and Computation (JCTC) and the Journal of Chemical Information and Modeling (JCIM), with Kitchin among the editorial authors. It asks what data, code and documentation authors of computational chemistry papers should be expected to share so that the principal findings can be checked, reproduced and extended, building on an earlier 2021 JCIM editorial on data and method availability (Amaro et al. 2026). The editors argue the need has grown because computational data sets are larger, workflows are more complex, and AI and ML are increasingly part of chemical research.
Approach
The editorial sets out a policy and a list of recommendations, framed by the FAIR principles (findable, accessible, interoperable, reusable). It is a policy statement, not a research study, so there are no results in the usual sense.
Key findings
The central policy is that, from May 1, 2026, the data and software needed to reproduce the key results of original research manuscripts should be available at submission whenever possible; Reviews, Perspectives and Editorials are exempt. Research articles should include a Data and Software Availability Statement saying where data, code and supporting material can be found, preferably in machine-readable form and in repositories that issue persistent identifiers. A template statement is provided.
The editors list eight kinds of material to share as appropriate: scripts and source code (with a statement of how the software can be obtained and under what license); representative input and output files with a clear repository layout and a README; system coordinates and custom parameters such as nonstandard basis sets or force-field changes; key raw numerical data such as energies and free energies; restart files where they matter; trajectories or representative structures for molecular dynamics; AI/ML models with architecture, hyperparameters, training, validation and test data, and the scripts or metadata needed to reproduce results, with the same data split enforced when methods are compared; and new potential energy surfaces together with the software needed to use them.
Manuscripts should state whether software is commercial, open source or in-house, give the exact version, and describe dependencies, runtime environment, hardware where relevant, and licensing limits. Where confidentiality, licensing or third-party ownership prevent full release, the main claims should still be reproducible from detailed methods and the unrestricted material, and confidential components should not be indispensable. Unreleasable code may be made temporarily available to reviewers, ideally preserving their anonymity. Work on quantum computing hardware or specialized platforms should report the hardware and software environment. Papers introducing new theory should give complete descriptions of the theory, algorithm and implementation, with key derivations in the paper or Supporting Information. Repositories run by sustainable institutions with long-term preservation commitments and DOIs or accession numbers are preferred.
Significance
For the Kitchin group, the editorial turns into journal policy positions Kitchin argued for a decade earlier: sharing machine-readable data and the code behind results (kitchin2015, kitchin2016a) and finding better ways to report machine-learned models (kitchin2018). The editors acknowledge that compliance takes effort and that best practices will evolve, and present the policy as building a culture of openness rather than imposing unnecessary burdens.