CoChat
Sign Up
Cited by 289Machine Learning in Materials ScienceAdvanced Chemical Physics StudiesComputational Drug Discovery Methods

Resolving Transition Metal Chemical Space: Feature Selection for Machine Learning and Structure–Property Relationships

Open access
Janet, Jon Paul, Kulik, Heather J.·2017·The Journal of Physical Chemistry A
DOI 10.1021/acs.jpca.7b08750
Open this paper in CoChat — free accountOpen PDF

Abstract

Machine learning (ML) of quantum mechanical properties shows promise for accelerating chemical discovery. For transition metal chemistry where accurate calculations are computationally costly and available training data sets are small, the molecular representation becomes a critical ingredient in ML model predictive accuracy. We introduce a series of revised autocorrelation functions (RACs) that encode relationships of the heuristic atomic properties (e.g., size, connectivity, and electronegativity) on a molecular graph. We alter the starting point, scope, and nature of the quantities evaluated in standard ACs to make these RACs amenable to inorganic chemistry. On an organic molecule set, we first demonstrate superior standard AC performance to other presently available topological descriptors for ML model training, with mean unsigned errors (MUEs) for atomization energies on set-aside test molecules as low as 6 kcal/mol. For inorganic chemistry, our RACs yield 1 kcal/mol ML MUEs on set-aside test molecules in spin-state splitting in comparison to 15-20× higher errors for feature sets that encode whole-molecule structural information. Systematic feature selection methods including univariate filtering, recursive feature elimination, and direct optimization (e.g., random forest and LASSO) are compared. Random-forest- or LASSO-selected subsets 4-5× smaller than the full RAC set produce sub- to 1 kcal/mol spin-splitting MUEs, with good transferability to metal-ligand bond length prediction (0.004-5 Å MUE) and redox potential on a smaller data set (0.2-0.3 eV MUE). Evaluation of feature selection results across property sets reveals the relative importance of local, electronic descriptors (e.g., electronegativity, atomic number) in spin-splitting and distal, steric effects in redox potential and bond lengths.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1021/acs.jpca.7b08750

Cites

117 of 117 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • By applying random forest-ranked recursive feature addition to KRR models, expected heuristics have been reinforced by feature selection approaches, such as the spin splitting of a complex being strongly dependent on the metal-local electronic environment.

    “By applying random forest-ranked recursive feature addition to KRR models, expected heuristics have been reinforced by feature selection approaches (22,32,105), such as the spin splitting of a complex being strongly dependent on the metal-local electronic environment (e.g., from the electronegativities of surrounding atoms; Figure 5)”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [32]

    Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

    Experimental — accuracy not yet measured

  • Low-cost, topological, three-dimensional, one-hot chemical encoding, and geometric descriptors are often employed for this purpose in transition-metal chemistry.

    “Low-cost, topological (5, 7, 32, 92-94), three-dimensional (95-100), one-hot chemical encoding (92, 101), and geometric (e.g., for MOF pores) (102-104) descriptors are often employed for this purpose in transition-metal chemistry.”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [32]

    Section 2b (p. 8) describes topological descriptors for transition-metal complexes and discrete metal and connecting-atom identities, supporting the topological and chemical-identity portions attributed to this paper. Our copy does not identify those identities as one-hot encoded. The sentence cites other sources for three-dimensional and geometric descriptors, so those categories are not attributed to this paper. (LLM, conf=0.82)

    Experimental — accuracy not yet measured

  • Identifying key descriptors or features in a structure-property relationship mapping has been demonstrated as an approach to revealing design principles and enabling iterative design.

    “Identifying key descriptors or features in a structure-property relationship mapping has been demonstrated as an approach to revealing design principles (18,22,32,52) and enabling iterative design (70).”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [32]

  • ML models, analysis of their key features, and dimensionality reduction represent important tools for building structure-property relationships and mapping chemical space.

    “ML models, analysis of their key features (32,52), and dimensionality reduction (i.e., unsupervised learning) (53-56) represent important tools for building structure-property relationships and mapping chemical space (57).”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [32]

  • One-shot models such as LASSO or random forests are widely used both as ML models for property prediction and for selection of important features.

    “In a similar fashion, one-shot models such as least absolute shrinkage and selection operator (LASSO) or random forests are widely used both as ML models for property prediction (61) and for selection of important features (22,32,40,43) (Figure 4).”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [32]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 4

  1. [1]1. Curtarolo, S.; Hart, G. L.; Nardelli, M. B.; Mingo, N.; Sanvito, S.; Levy, O., The HighThroughput Highway to Computational Materials Design. Nature materials 2013, 12 , 191-201.
  2. [68]68. Puzyn, T.; Leszczynski, J.; Cronin, M. T., Recent Advances in QSAR Studies: Methods and Applications . Springer Science & Business Media: 2010; Vol. 8.
  3. [74]74. Faber, F. A.; Hutchison, L.; Huang, B.; Gilmer, J.; Schoenholz, S. S.; Dahl, G. E.; Vinyals, O.; Kearnes, S.; Riley, P. F.; Anatole von Lilienfeld, O., Machine Learning Prediction Errors Better Than DFT Accuracy. ArXiv e-prints 2017, 1702 , arXiv:1702.05532. · arXiv
  4. [82]82. Guyon, I. E., André An Introduction to Variable and Feature Selection. Journal of Machine Learning Research 2003, 3 , 1157-1182.

Cited by

Sorted by citations

Related

Authors
Janet, J. P., & Kulik, H. J.
Affiliations
Massachusetts Institute of Technology (US)
Corresponding author
Kulik, Heather J.
ORCID
Venue
The Journal of Physical Chemistry A
Publisher
American Chemical Society (ACS)
Published
2017-11-15
Volume
121
Issue
46
Pages
8939-8954
ISSN
  • 1089-5639 (ISSN-L)
  • 1520-5215
DOI
10.1021/acs.jpca.7b08750
OpenAlex
W2749580687
Cited by
289
References
117
Citations by year
2026: 34 · 2025: 35 · 2024: 48 · 2023: 37 · 2022: 38 · 2021: 28 · 2020: 31 · 2019: 17 · 2018: 20 · 2017: 1
FWCI
7.97
Type
journal article
Language
en
Open access
Locations
  • The Journal of Physical Chemistry A (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc)
  • arXiv (Cornell University)
Funders
  • Massachusetts Institute of Technology
  • Division of Electrical, Communications and Cyber Systems (ECCS-1449291)
  • Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET-1704266)
  • Burroughs Wellcome Fund
Topics
  • Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
  • Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Advanced Chemical Physics Studies
  • Physical Sciences › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
Keywords
transition metal chemistry, autocorrelation function, feature selection, structure-property relationships, redox potential, metal-ligand bond lengths, atomization energies, random forest, LASSO, recursive feature elimination, inorganic molecules

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review