comment

This	is	the	peer	reviewed	version	of	the	following	ar3cle:	Nature	Catalysis	2018,	1,	809-810,	which	has	been	published	in	final	form	at	hDps://www.nature.com/
ar3cles/s41929-018-0176-4.	This	ar3cle	may	be	used	for	non-commercial	purposes	in	accordance	with	Macmillan	Terms	and	Condi3ons	for	Self-Archiving.

The role of computational results databases in
accelerating the discovery of catalysts
Databases of computational results hold high promise for accelerating catalysis research. Still, many challenges
remain and consensus on facets such as metadata, reliability and curation is crucial to transform the hype into an
attractive technology.

Carles Bo, Feliu Maseras and Núria López

S

cience has the power to generate
great volumes of data. Over the years,
numerous collective efforts, including
those made by the Manhattan Project,
CERN, the Genomics Consortium and
large research facilities in astronomy and
climate, have had to collect, label, store and
curate scientific digital outputs. Chemistry
and materials science are different. Highly
standardized thermodynamic data have
been curated by organizations such as
the National Institute of Standards and
Technology (NIST) through Chemistry
Handbooks1, but probably the most
successful stories in chemistry-related
data archiving are the neat curation of a
protein database2 and the maintenance of
crystallographic records for molecules3
and materials4. In addition, private efforts
made by chemical companies ensure that
they can trace experiments performed
more than half a century ago.
Computational techniques employed
in the study of the atomistic processes
occurring at the sub-nanometre scale
in chemistry, biochemistry, physics and
materials science have been based on solving
the Schrödinger equation in different
ways. It has been calculated that about
30% of the total use of supercomputers at
the European level are devoted to various
kinds of density functional theory (DFT)
approximations. Now, the robustness of the
implementations in the different quantum
chemistry codes ensures that the data
obtained are of the same quality irrespective
of the computational codes, provided that
high standards are used5. Multipurpose,
Dropbox-like initiatives such as Zenodo,
Figshare and Dryad allow the allocation
of space for storage of scientific data.
However, the massive simulations generate
high volumes of data that are neither
tagged nor syndicated, thus the efforts of
computation are partially in vain.
Several initiatives have been developed
to systematically collect information from
calculations and build databases that can be
NATURE CATALYSIS | www.nature.com/natcatal

easily mined and can serve as the first step
in artificial intelligence models. One of the
earliest examples is the Quantum Chemistry
Literature Database, which collected data
from published manuscripts6. This evolved
into a website7, but updates seem to have
been discontinued in 2013. Some reference
databases such as the Computational
Chemistry Comparison and Benchmark
Database by NIST8, the Benchmark Energy
and Geometry Database9 or the Minnesota
Databases10 boosted the development of
new DFT functionals, solvation models
and empirical dispersion corrections.
Others focus on specific issues, such as the
Alexandria Library, dedicated to force field
development11 and the PubChemQC project,
which took data encoded in the PubChem
database and calculated the first 10 excited
states for over 2 million molecules12. The
software generates different kinds of inputs
for molecular codes using Open Babel13.
For instance, it can search for molecules
where the HOMO–LUMO gap is smaller
than a certain value (for example, 1.0 eV).
Regarding structured data, the pioneering
definition of the Chemical Markup
Language14 was followed by The Quixote
Project15, which put forward an ambitious
programme for collaborative and open
quantum chemistry data management.
In turn, CatApp16 was among the first
attempts to recycle computational data in
the field of heterogeneous catalysis.
Nowadays, stand-alone data extractors,
such as ExcelAutomat17, EsiGen18 and the
cclib library19 are available. They search
in output files for patterns, which are
then parsed, summarized, uploaded to
spreadsheets and/or assigned to variables
for molecular codes. In addition, data
repositories and specialized software
platforms are currently under development.
But it is in materials science where the
developments have arguably been the
most exciting. For instance the US-driven
Materials Genome Initiative20 has a
specific Materials Project21, which aims at

employing supercomputers and the most
advanced electronic structure analyses to
provide an open web-based access to all
the structures already computed (including
those not yet synthetized), and the resources
needed for the design of novel materials22.
Important efforts have also been made at
the European level through the European
Materials Modelling Council23 to achieve
transferability of data and codes through the
Modelling Data conceptual frameworks24.
Although the core of the work has been
related to transferability in multiscale
modelling (data inheritance between scales),
some excellent work on ontologies has
been developed. In Europe, two initiatives
have been working in parallel as centres
of excellence (2014–2017), namely the
Novel Materials Discovery (NoMaD)
repository25 and the Automated Interactive
Infrastructure and Database (AiiDA)26,27.
NoMaD creates, collects, stores and cleanses
computational materials science data, and
develops tools for data mining to find
structures, correlations and novel results
that do not appear within smaller datasets.
The system is decentralized and can
incorporate results from different sources.
Particular areas of interest are heat-transport
tensors for many materials, catalytic
activation of CO2 and thin coating films to
protect novel hybrid perovskite solar cells
from degradation in moist environments
through high-throughput screening of
potential transparent oxide semiconductors.
In turn, AiiDA is a flexible and scalable
infrastructure that allows the management,
preservation and dissemination of
computational results, but also works on the
data ensuring its searchability and providing
workflows. The core of their set-up is driven
by automation, data, environment and
sharing, adopting concepts and tools from
computer science. Particularly, the Materials
Cloud28 has been developed to improve the
viability of industrially adapted databases.
However, the changes in the European
policies with the new Virtual Materials

comment
Market Place strategy might affect these
initiatives. The Computational Materials
Repository29 is structured in different
projects — for instance two-dimensional
materials — and extracts the data through
the combined use of the Atomic Simulation
Environment (ASE) and Python scripts. In
turn, ioChem-BD (Input/Output Chemistry
Big Data)30,31 focuses on the chemical
properties and links data generation and
open-access publication. In addition to
providing tools for data curation and
post-processing, ioChem-BD builds up
a distributed network of independent
nodes around a central server where
data and metadata are indexed. In any
of these databases, the transferability of
the information between the molecular
and periodic approaches is a challenge,
particularly when it comes to addressing
chemical and catalytic problems where the
recognition of the structural and electronic
patterns that build the active site is a
major goal. Merging the best properties of
homogeneous and heterogeneous catalysts,
for instance in the area of single-atom
catalysis32, requires this pattern recognition.
However, the representation of crystalline
materials in a format that allows the
comparison to their molecular counterparts
is a major hurdle33.
While the path is well traced, efforts are
still needed to give the proper semantics
to data34 in order to allow transferability
between the different fields where firstprinciples results can be relevant. Among
the challenges ahead, it is mandatory to
transform data available in plain websites
into structured data, accompanied by
valuable metadata, which follows the
findable, accessible, interoperable, recyclable
(FAIR) principles. Code interoperability
and standard data formats are motivating
international research actions35,36, such as
the Molecular Sciences Software Institute37
in the US and the European Materials
Modelling Consortium24, but much effort
worldwide is still needed. This will allow
the integration in multiscale methodologies
that can go from the atomistic perspective
to the device level. The integration with

experimental results is yet another longterm challenge. Finally, the combination of
databases with machine learning is having
a big impact on the field by allowing an
increasing number of applications33,38–42 and
is seen as a major opportunity in industry43.
However, this blooming computational field
also demands massive curated data and
robust algorithm benchmarks44, as there is a
risk of the hype masking the real advantages
of these powerful tools45,46.
All of these advances will ensure that
artificial intelligence technologies based on
computational or mixed computational/
experimental databases will accelerate the
discovery of active, selective and stable
catalysts that, for instance, are able to break
the standard linear scaling relationships,
follow the principles of green chemistry and
provide the tools for a circular economy
in the areas of chemistry, catalysis and
materials science.
❐
Carles Bo1,2, Feliu Maseras1,3 and
Núria López1*

Institute of Chemical Research of Catalonia,
The Barcelona Institute of Science and Technology,
Tarragona, Spain. 2Universitat Rovira i Virgili,
Departament de Química Física i Inorgànica,
Tarragona, Spain. 3Universitat Autònoma
de Barcelona, Departament de Química,
Bellaterra, Spain.
*e-mail: nlopez@iciq.es

1

Published: xx xx xxxx
https://doi.org/10.1038/s41929-018-0176-4
References

1. NIST Chemistry WebBook. NIST Standard Reference Database
Number 69 (NIST, accessed 26 May 2018); https://webbook.nist.
gov/chemistry
2. Protein Data Bank (RCSB, accessed 26 May 2018); https://www.
rcsb.org
3. The Cambridge Structural Database (CCDC, accessed
26 May 2018); https://www.ccdc.cam.ac.uk
4. Inorganic Crystal Structure Database (FIZ Karlsruhe, accessed
26 May 2018); http://www2.fiz-karlsruhe.de/icsd_home.html
5. Lejaeghere, K. et al. Science 351, aad3000 (2016).
6. Ohno, K. & Morokuma, K. Quantum Chemistry Literature Data
Base—Bibliography of Ab Initio Calculations for 1978–1980
(Elsevier, Amsterdam, 1982).
7. QCLDB II (QCDB Group, accessed 25 May 2018); http://qcldb2.
ims.ac.jp
8. Computational Chemistry Comparison and Benchmark Database,
NIST Standard Reference Database Number 101 Release 19 (NIST,
accessed 26 May 2018); https://cccbdb.nist.gov

9. Hobza, P. Benchmark Energy and Geometry Database
(Institute of Organic Chemistry and Biochemistry, Prague,
accessed 26 May 2018); http://www.begdb.com
10. Databases Truhlar Research Group (accessed 26 May 2018);
http://truhlar.chem.umn.edu/content/databases
11. Ghahremanpour, M. M., van Maaren, P. J. & van der Spoel, D.
Sci. Data 5, 180062 (2018).
12. Nakata, M. & Shimazaki, T. J. Chem. Inf. Model 57,
1300–1308 (2017).
13. Open Babel: The Open Source Chemistry Toolbox (accessed
26 May 2018); http://openbabel.org/wiki/Main_Page
14. Murray-Rust, P. & Rzepa, H. S. J. Cheminformatics 3, 44 (2011).
15. Adams, S. et al. J. Cheminformatics 3, 38 (2011).
16. Hummelshøj, J. S., Abild-Pedersen, F., Studt, F., Bligaard, T. &
Nørskov, J. K. Angew. Chem. Int. Ed. 51, 272–274 (2011).
17. Laloo, J. Z. A., Laloo, N., Rhyman, L. & Ramassami, P.
J. Comput. Aided Mol. Des. 31, 667–673 (2017).
18. Rodríguez-Guerra Pedregal, J., Gómez-Orellana, P. &
Maréchal, J.-D. J. Chem. Inf. Model. 58, 561–564 (2018).
19. O’Boyle, N. M., Tenderholt, A. L. & Langner, K. M.
J. Comput. Chem. 29, 839–845 (2008).
20. Materials Genome Initiative (accessed 29 May 2018);
https://mgi.gov
21. The Materials Project (accessed 30 August 2018); https://www.
materialsproject.org
22. Tabor, D. P. et al. Nat. Rev. Mater. 3, 5 (2018).
23. The European Materials Modelling Council (accessed
30 August 2018); https://emmc.info
24. de Bass, A. F. What Makes a Material Function (EU, 2017)
25. NOMAD Repository (NOMAD Laboratory, accessed 23 May
2018); https://nomad-repository.eu
26. Automated Interactive Infrastructure and Database for Computational
Science (accessed 25 May 2018); http://www.aiida.net
27. Pizzi, G., Cepellotti, A., Sabatini, R., Marzari, N. & Kozinsky, B.
Comput. Mater. Sci. 111, 218–230 (2016).
28. Web platform “Materials Cloud” could help industry streamline
research efforts. Marvel http://nccr-marvel.ch/highlights/201805-web-platform-materials-cloud-could-help-industry
(30 May 2018).
29. Computational Materials Repository (CAMd, accessed 14
September 2018); https://cmr.fysik.dtu.dk
30. Álvarez-Moreno, M. et al. J. Chem. Inf. Model. 55, 95 (2015).
31. ioChem-BD (accessed 29 May 2018); http://www.iochem-bd.org
32. Chen, Z. Nat. Nanotech 13, 702–707 (2018).
33. Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O. &
Walsh, A. Nature 559, 547–555 (2018).
34. Wang, B., Dobosh, P. A., Chalk, S., Sopek, M. & Ostlund, N. S.
J. Phys. Chem. A 121, 298–307 (2016).
35. Rossi, E. et al. J. Comput. Chem. 35, 611–621 (2014).
36. Ghiringhelli, L. M. npj Comput. Mater. 3, 46 (2017).
37. The Molecular Sciences Software Institute (accessed 30 August
2018); https://molssi.org
38. Schütt, K. T., Arbabzadah, F., Chmiela, S. & Müller, K.-R.
Nat. Commun. 8, 13890 (2017).
39. Janet, J. P. & Kulik, H. J. Chem. Sci. 8, 5137–5152 (2017).
40. Ferguson, A. L. ACS Cent. Sci. 4, 938–941 (2018).
41. Gómez-Bombarelli, R. et al. ACS Cent. Sci. 4, 268–276 (2018).
42. Nandy, A., Duan, C., Janet, J. P., Gugler, S. & Kulik, H. Preprint at
https://doi.org/10.26434/chemrxiv.6987074.v1 (2018).
43. Jones, G. Nat. Catal. 1, 311–313 (2018).
44. Wu, Z. et al. Chem. Sci. 9, 513–530 (2018).
45. Lemonick, S. Is machine learning overhyped? Chem. Eng. News
https://cen.acs.org/physical-chemistry/computational-chemistry/
machine-learning-overhyped/96/i34 (2018).
46. PASC18 panel discussion. Is HPC facing a game change?
YouTube https://www.youtube.com/watch?v=mTqzCvm0G5c
(16 July 2018).

NATURE CATALYSIS | www.nature.com/natcatal