Pith. sign in

REVIEW 4 major objections 6 minor 47 references

3D Molecular Representation Learning for Organic Mixtures: Viscosity and Density Prediction

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fine-tuned 3D molecular encoder predicts viscosity and density of organic mixtures with test-set R2 of 0.973 and 0.996, outperforming descriptor-based baselines and capturing non-monotonic mixing behavior.

desk verdict A credible and practically useful mixture-property predictor built on Uni-Mol, held up by a strong grouped split and independent lab checks, but the split's component-order canonicalization must be clarified before the headline R2 is trusted. read the letter →

arxiv 2608.06697 v1 pith:23PIZYVC submitted 2026-08-07 physics.chem-ph

Haicheng Qu , Yanyi Su , Ning Wang , Shangqian Chen , Zhifeng Gao , Jun Cheng , Qi Ou This is my paper · ORCID
classification physics.chem-ph
keywords 3DmolecularrepresentationlearningorganicmixturesviscositypredictiondensityUni-Molfine-tuningmixturepropertyhydrogenbondingternary
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper introduces Uni-Mix, a machine-learning model that predicts the dynamic viscosity and density of organic mixtures from molecular structure, composition, and temperature. The authors' central claim is that a 3D molecular encoder pre-trained on 209 million conformations, when fine-tuned on roughly 29,000 binary-mixture viscosity records and 25,000 density records from the Landolt–Börnstein database, achieves held-out test-set R2 values of 0.973 (log10 viscosity) and 0.996 (density). If this claim holds, formulators of lubricants, solvents, and heat-transfer fluids gain a fast alternative to exhaustive experiments and to descriptor-based models that miss cooperative intermolecular interactions. The model also reproduces a non-monotonic viscosity maximum in water–acetic acid mixtures that simple mixing rules cannot capture, and preliminary results suggest the same architecture transfers to ternary hydrocarbon mixtures.

What carries the argument

The load-bearing mechanism is the mixture interaction module: stacked multi-head self-attention layers operating on per-molecule 3D embeddings, with temperature encoded by sinusoidal positional encoding and mole fractions injected by element-wise multiplication, followed by aggregation to a 256-dimensional mixture representation. This attention module acts as a learned surrogate for intermolecular interactions, letting multiple components share one encoder and communicate before the final property readout, while a masking mechanism lets the same architecture handle pure components and mixtures.

What would settle it

Hold out all mixture pairs containing a carboxylic acid or an alcohol from training, retrain, and test on those held-out pairs: if test-set R2 for these hydrogen-bonding systems falls well below the reported 0.973, the model's accuracy depends on having seen chemically similar pairs during training rather than on a transferable representation of interaction physics.

Watch

Extended reading notes

Core claim

Uni-Mix encodes each component as a single RDKit-generated 3D conformer passed through the shared pre-trained Uni-Mol encoder to obtain a 512-dimensional embedding, concatenates 15 RDKit descriptors, then applies multi-head self-attention over the component embeddings, with temperature injected through sinusoidal positional encoding and mole fractions through element-wise multiplication. Fine-tuned with a Gaussian negative log-likelihood loss, the model yields test R2 of 0.973 for dynamic viscosity (MAE 0.222 cP, RMSE 0.890 cP in native space) and 0.996 for density (MAE 0.006 g/cm3), outperforming SVM, random forest, and XGBoost on identical splits. The paper further claims that the model captures non-monotonic viscosity-composition curves in hydrogen-bonding systems, such as water–acetic acid, and that a general model fine-tuned across chemical families transfers better to ester/anhydride systems than a specialized model, while matching it for hydrocarbons. On two ternary hydrocarbon mixtures never seen in training, test R2 values of 0.928 for viscosity and 0.944 for density over 12 points are reported.

Load-bearing premise

The accuracy claim rests on the assumption that one static RDKit-generated 3D conformer per molecule, combined with a generic self-attention interaction module, can represent the hydrogen-bonding, dipole, and packing interactions that govern non-ideal mixture viscosity.

Editorial extensions

If this is right

  • If the reported accuracy reproduces outside the Landolt–Börnstein corpus, Uni-Mix can screen organic mixture formulations for viscosity and density without new experiments, constraining lubricant and solvent design space.
  • The architecture does not assume a fixed number of components, so the same fine-tuned encoder can be extended to quaternary and higher mixtures; the ternary results are the first evidence of that transfer.
  • Because the model learns non-monotonic composition dependence, it can be used to identify compositions where mixture viscosity exceeds that of either pure component, as near x_acid ≈ 0.5–0.6 in water–acetic acid.
  • The structural analysis implies quantitative guidelines for viscosity-index tuning: multi-ring naphthenic and aromatic components degrade the ν40/ν100 ratio more than isoparaffins or alkylbenzenes, which could direct base-oil formulation.
  • The general model is the safer default for polar functional-group families such as esters and anhydrides, where specialized models underperform because of limited within-family training data.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: The paper's reliance on a single RDKit conformer per molecule suggests a natural stress test: generate an ensemble of low-energy conformers and check whether prediction variance is large; if it is, conformational sampling, not just mixture attention, is the next limiting factor.
  • Inference: The residual error concentrated in carboxylic-acid systems leaves open the possibility that the model is interpolating within chemically similar pairs rather than learning hydrogen-bond topologies; a leave-class-out evaluation on all acid-containing pairs would distinguish these.
  • Inference: The ternary test is limited to 12 points of straight-chain hydrocarbons, but the architecture's permutation-invariant component encoding suggests it could be extended to arbitrary multicomponent systems; the strongest test would be a ternary mixture containing an alcohol or acid, where non-ideal interactions are strongest.
  • Inference: The viscosity-temperature sensitivity analysis (ν40/ν100) is computed from model predictions, not measurements, so the ranking of the five hydrocarbon structures is a model prediction that could be checked experimentally on a small set of PAO blends.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper introduces Uni-Mix, a mixture-aware 3D molecular representation learning model built on the pre-trained Uni-Mol encoder, for predicting the dynamic viscosity and density of organic mixtures. Each component is represented by a single RDKit-generated 3D conformer embedded by Uni-Mol, concatenated with 15 RDKit descriptors; a multi-head self-attention module processes the component embeddings, with mole fractions and temperature injected as input features. The model is fine-tuned with a Gaussian NLL loss on Landolt–Börnstein binary organic mixture data, achieving test-set R2 of 0.973 for viscosity (in log10 space) and 0.996 for density. The paper compares against SVM, random forest, and XGBoost baselines, tests general versus specialized models on hydrocarbon and ester/anhydride subsets, validates on independently measured samples, presents a 12-point ternary proof-of-concept, analyzes hydrogen-bonding system types, and uses the model to predict how hydrocarbon structure affects the viscosity-temperature index.

Significance. If the results hold, Uni-Mix is a practical and transferable tool for mixture property prediction, with clear improvements over descriptor-based baselines and an architecture that naturally extends to multicomponent systems. The paper has notable strengths: a pair-grouped split that guards against pair leakage if implemented order-insensitively, independent experimental validation on hydrocarbon and ester systems, a ternary demonstration, a public model URL, and a promise of data availability (data.zip with PM6-optimized coordinates). The structure-effect analysis in Figure 7 constitutes falsifiable predictions for future experiments. At the same time, the central empirical claim rests on a single random split without error bars, Table 5 reports in-sample metrics, and the water–acetic acid example does not disclose whether it is a held-out system; these details currently limit the strength of the claims. The representation also uses a single conformer per molecule, and the paper itself notes residual outliers in carboxylic-acid systems, so the physical generality for strongly interacting mixtures remains an open question.

major comments (4)
  1. [Methods, pair-grouped split; SI 'Data pre-processing and split strategy'] The grouped split is keyed on the ordered tuple (SMILES_I, SMILES_II). For a binary mixture, (A,B) and (B,A) with complementary mole fractions are the same physical system. If the source database lists components in different orders across records, records of one physical pair can be split between training and test sets, inflating the reported test R2 values. The Methods and SI never state that SMILES are canonicalized or sorted before grouping, and no split script is released to verify this. Please confirm that pair grouping is order-insensitive, report the number of reversed-pair records in the raw data, or release the grouping code. If canonicalization is confirmed, this concern is resolved.
  2. [Table 5 and 'Impact of hydrogen bonding'] Table 5 reports R2 and RMSE over the complete dataset (training and test sets combined), and the text states that 'the model still achieves satisfactory accuracy across all system types (R2 ≥ 0.803)'. Because these statistics include training data, they do not demonstrate held-out generalization for the hydrogen-bonding categories. Please recompute the system-type breakdown on the test set only, or relabel Table 5 explicitly as an in-sample fit.
  3. [Figure 6 and water-acetic acid example] The text does not disclose whether the water–acetic acid mixture was part of the training set, the test set, or an external validation set. Since capturing non-monotonic viscosity-composition behavior is a central qualitative claim, please state the split membership (or that the system lies outside the Landolt–Börnstein corpus). If this pair was in training, the figure demonstrates interpolation rather than generalization.
  4. [Methods, split paragraph; Table 2] All headline metrics and baseline comparisons are obtained from a single grouped random split, with no repeated splits, confidence intervals, or standard deviations. Given that the viscosity R2 gap over XGBoost is 0.973 versus 0.955, the reader cannot assess whether the improvement is statistically meaningful. Please report metrics over multiple grouped splits (for example, five random seeds) or provide confidence intervals for the test-set metrics.
minor comments (6)
  1. [Methods, temperature encoding equation] The equation for the sinusoidal temperature encoding is garbled in the text ('PE(T)!=sin*Tω!/$,...'); please write it as PE(T,2k)=sin(T/ω^{2k/d}) and PE(T,2k+1)=cos(T/ω^{2k/d}) (or the intended form) and define all symbols precisely.
  2. [Methods, temperature encoding] The encoding is described as 'parameter-free'; while it introduces no trainable parameters, the base frequency ω and dimension d are hyperparameters chosen by the authors. Suggest rewording to 'without additional trainable parameters'.
  3. [Table 5 note] Table 5 evaluates R2 in the native linear space for viscosity, whereas the rest of the paper evaluates viscosity R2 in log10 space; this is disclosed in the note, but it makes those R2 values not directly comparable with Table 2. Please state the rationale more prominently or report both spaces.
  4. [Results, comparison with Ref. 2] The sentence 'the MAE in log10(cP) given by our model is 0.035, surpassing the 0.043 reported in a recent GNN-based study' compares across different datasets and splits; please qualify this as an informal reference point rather than a controlled benchmark.
  5. [Data and code availability; SI] The main text says the datasets are 'obtained upon reasonable request', while the SI states 'the original training and test data sets ... are compressed in the supplementary file data.zip'. Please harmonize these statements and ensure the data file is actually provided or a stable repository link is given, so that the split-level checks requested above can be performed by readers.
  6. [Figure 7 / Table S16] The ν40/ν100 analysis in Figure 7 is based entirely on model predictions without experimental verification; the main text should state this explicitly, since Table S16 is a prediction table rather than a measurement table.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the reported test-set R2 values are held-out against pair-grouped split data, and the self-cited Uni-Mol backbone is independent pretrained infrastructure rather than an encoding of the target mixture viscosities.

full rationale

The central quantitative claims are the test-set R2 values of 0.973 for dynamic viscosity and 0.996 for density, obtained by fine-tuning a Uni-Mol-based encoder on Landolt-Börnstein binary mixture data and evaluating on a grouped split keyed on the molecular pair. No parameter is fitted to the test set, and the SI explicitly states that 'all records sharing the same molecular pair—across all mole fractions and temperatures—are assigned as a unit,' so the headline numbers are external to the fit rather than constructed from it. The self-citations to Uni-Mol (ref. 35) and Uni-ELF (ref. 39) underpin the architecture but are not load-bearing in the derivation: Uni-Mol is an open pretrained model with a separate published objective on 209 million conformations, not on the present viscosity/density targets, so the citation is independent evidence for the encoder rather than a restatement of the output. The structure-effect analysis uses model predictions as a surrogate for experiment, which is a scope limitation, not circular reasoning. Two non-circular caveats should be flagged: the SI describes the split as 'grouped by the SMILES pair' without explicitly stating that component order is canonicalized, so reversed-pair leakage cannot be fully ruled out from the text; and the Data and code availability section states that 'Complete fine-tuning datasets can be obtained upon reasonable request,' with no code release described. Both are data-integrity and reproducibility concerns, not reductions of the prediction to its inputs. Overall, the derivation chain is self-contained for the claims made.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The quantitative claims rest on a large fitted parameter set plus assumptions about the transferability of the pretrained 3D encoder, the adequacy of single RDKit conformers, the accuracy of the Landolt-Börnstein compilation, and the ability of attention over per-molecule embeddings to capture intermolecular physics. No fundamentally new physical entity is introduced.

free parameters (4)
  • Fine-tuned network weights (encoder plus mixture interaction module)
    The central claim depends on millions of parameters optimized on the Landolt-Börnstein viscosity and density data via Gaussian NLL loss; the reported R2 and MAE are properties of this fitted model.
  • Fine-tuning hyperparameters = learning rate 4e-5, batch size 32, 4 attention layers, 8 heads, embedding dimension 256, early stopping patience 30…
    These are chosen by hand and early stopping rather than derived; different choices could change the test metrics.
  • Temperature sinusoidal encoding base frequency and dimension = omega = 10000, d = 256
    Fixed encoding hyperparameters that shape the temperature input; no ablation is provided to show they are optimal.
  • Gaussian NLL variance clamp range = [-10, 10]
    Chosen for numerical stability; it affects how strongly difficult samples are down-weighted in the loss.
assumptions (6)
  • domain assumption Uni-Mol pretrained on 209M conformations provides transferable 3D molecular representations for target properties
    The method inherits the encoder without retraining on mixture data; if pretrained embeddings lack relevant mixture interaction information, predictions degrade.
  • domain assumption RDKit-generated single conformers adequately represent molecular 3D structure for each component
    Methods state conformers are generated from SMILES via RDKit, while the SI mentions PM6-optimized coordinates; single conformers also ignore conformational ensembles and solvent environment.
  • domain assumption Landolt-Börnstein compiled experimental data are accurate and homogeneous enough for training
    All metrics are evaluated against these literature values; any compilation errors propagate into the reported R2 and MAE.
  • ad hoc to paper Self-attention over per-molecule embeddings, with mole fractions and temperature injected as input features, can represent composition- and temperature-dependent intermolecular interactions
    The architecture imposes no explicit intermolecular geometry or physics; the ability to capture non-ideal mixing is an empirical claim, not a guaranteed property.
  • domain assumption Training/test split grouped by molecular pair prevents pair-level leakage, and the single random 8:2 split is representative
    A single random split is used; no cross-validation or uncertainty quantification across splits is provided.
  • ad hoc to paper Sinusoidal temperature encoding with base frequency 10000 is sufficient to capture temperature dependence
    No ablation shows that this encoding is preferable to alternatives; the functional form is chosen for convenience.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3D Molecular Representation Learning for Organic Mixtures: Viscosity and Density Prediction." pith.science (2026). https://pith.science/paper/23PIZYVC

@misc{pith2026260806697,
  author       = {Pith},
  title        = {Pith review of: 3D Molecular Representation Learning for Organic Mixtures: Viscosity and Density Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/23PIZYVC}},
  note         = {Machine review of arXiv:2608.06697}
}
read the original abstract

The viscosity and density of organic mixtures are essential properties for designing lubricants, solvents, and heat transfer fluids. In engineering practice, formulating a functional fluid requires understanding how these properties change with composition and temperature. However, exhaustive experimental characterization across the full parameter space is impractical due to the vast number of possible species and combinations. Here we introduce a mixture-aware 3D molecular representation learning strategy, built upon a pre-trained molecular encoder, that jointly encodes component structures, mole fractions, and temperature to achieve accurate predictions for organic mixtures. Fine-tuning on publicly available datasets covering a wide range of binary organic mixtures yields test-set R2 values of 0.973 for dynamic viscosity and 0.996 for density, significantly outperforming traditional machine learning baselines. Beyond this overall accuracy, the model captures non-monotonic viscosity changes upon mixing, surpassing simple linear or logarithmic mixing rules. The architecture is extendable to ternary and multicomponent mixtures, as verified via preliminary experiments. Using this model, we quantitatively analyze how molecular structure-branching, cycloalkane, and aromatic rings-affects viscosity-temperature behavior, which benefits the design of lubricants with superior viscosity-temperature performance. Altogether, this work provides a practical, data-driven tool for mixture property prediction, accelerating the rational formulation of functional fluids in chemical engineering.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 45 canonical work pages

  1. [1]

    Sarpa, D. et al. Computing Viscosities of Mixtures of Ester-Based Lubricants at Different Temperatures. J. Phys. Chem. B 127, 2587–2594 (2023)

    work page 2023

  2. [2]

    Bilodeau, C. et al. Machine learning for predicting the viscosity of binary liquid mixtures. Chemical Engineering Journal 464, 142454 (2023)

    work page 2023

  3. [3]

    & Martini, A

    Panwar, P., Yang, Q. & Martini, A. Temperature-Dependent Density and Viscosity Prediction for Hydrocarbons: Machine Learning and Molecular Dynamics Simulations. J. Chem. Inf. Model. 64, 2760–2774 (2024). Article

    work page 2024

  4. [4]

    Chew, A. K. et al. Leveraging high-throughput molecular simulations and machine learning for the design of chemical mixtures. npj Comput Mater 11, 72 (2025)

    work page 2025

  5. [5]

    Martini, A., Ramasamy, U. S. & Len, M. Review of Viscosity Modifier Lubricant Additives. Tribol Lett 66, 58 (2018)

    work page 2018

  6. [6]

    & Erdemir, A

    Holmberg, K. & Erdemir, A. Influence of tribology on global energy consumption, costs and emissions. Friction 5, 263–284 (2017)

    work page 2017

  7. [7]

    F., Sartori, E., Crepaldi, J

    Rovai, F. F., Sartori, E., Crepaldi, J. & Rajala, S. Engine Lubricant Impact in Light-Vehicle Fuel Economy: A Combined Numerical Simulation and Experimental Validation. Lubricants 13, 137 (2025)

    work page 2025

  8. [8]

    Recent advancements in volumetric flow meter for industrial application

    Mohindru, P. Recent advancements in volumetric flow meter for industrial application. Heat Mass Transfer 59, 2149–2166 (2023)

    work page 2023

Show all 47 references
  1. [10]

    & Vrabec, J

    Miroshnichenko, S. & Vrabec, J. Excess properties of non-ideal binary mixtures containing water, methanol and ethanol by molecular simulation. Journal of Molecular Liquids 212, 90–95 (2015)

    2015

  2. [11]

    Mahajan, A. R. & Mirgane, S. R. Excess Molar V olumes and Viscosities for the Binary Mixtures of n-Octane, n-Decane, n-Dodecane, and n-Tetradecane with Octan-2-ol at 298.15 K. Journal of Thermodynamics 2013, 1–11 (2013)

    2013

  3. [12]

    M., Lichtenthaler, R

    Prausnitz, J. M., Lichtenthaler, R. N. & De Azevedo, E. G. Molecular Thermodynamics of Fluid-Phase Equilibria. (Pearson Education, 1998)

    1998

  4. [13]

    J., Diamantonis, N

    Tillotson, M. J., Diamantonis, N. I., Buda, C., Bolton, L. W. & Müller, E. A. Molecular modelling of the thermophysical properties of fluids: expectations, limitations, gaps and opportunities. Phys. Chem. Chem. Phys. 25, 12607–12628 (2023)

    2023

  5. [14]

    Kirova, E. M. & Norman, G. E. Viscosity calculations at molecular dynamics simulations. J. Phys.: Conf. Ser. 653, 012106 (2015)

    2015

  6. [15]

    & Maginn, E

    Zhang, Y ., Otani, A. & Maginn, E. J. Reliable Viscosity Calculation from Equilibrium Molecular Dynamics Simulations: A Time Decomposition Method. J. Chem. Theory Comput. 11, 3537–3546 (2015)

    2015

  7. [16]

    & Das, R

    Roy, K., Kar, S. & Das, R. N. QSAR/QSPR Modeling: Introduction. in A Primer on QSAR/QSPR Modeling 1–36 (Springer International Publishing, Cham, 2015). doi:10.1007/978-3-319-17281-1_1

    2015 doi

  8. [17]

    Shi, D. et al. Deep insights into the viscosity of deep eutectic solvents by an XGBoost-based model plus SHapley Additive exPlanation. Phys. Chem. Chem. Phys. 24, 26029–26036 (2022)

    2022

  9. [18]

    Chew, A. K. et al. Advancing material property prediction: using physics-informed machine learning models for viscosity. J Cheminform 16, 31 (2024)

    2024

  10. [19]

    & Hemmati-Sarapardeh, A

    Rezaei, F., Akbari, M., Rafiei, Y . & Hemmati-Sarapardeh, A. Compositional modeling of gas-condensate viscosity using ensemble approach. Sci Rep 13, 9659 (2023)

    2023

  11. [20]

    & Dey, R

    Prabhune, A., Mathur, A., Saha, S. & Dey, R. Predictive, correlative and machine learning models for estimation of viscosity of liquid mixtures. Journal of Molecular Liquids 397, 124147 (2024). Article

    2024

  12. [21]

    & Bamler, R

    Zenn, J., Gond, D., Jirasek, F. & Bamler, R. Balancing molecular information and empirical data in the prediction of physico-chemical properties. Digital Discovery 4, 683–693 (2025)

    2025

  13. [22]

    Yue, X. et al. Density, dynamic viscosity, excess property and intermolecular interplay studies for 1,4-butanediol + dimethyl sulfoxide binary mixture. Journal of Molecular Liquids 263, 40–48 (2018)

    2018

  14. [23]

    & Vatanparast, M

    Almasi, M. & Vatanparast, M. Temperature dependent behavior of hydrogen-bonded liquids: bridging experiments with molecular dynamics and DFT. Sci Rep 16, 9185 (2026)

    2026

  15. [24]

    Huwaimel, B., Alanazi, J., Alanazi, M., Alharby, T. N. & Alshammari, F. Computational models based on machine learning and validation for predicting ionic liquids viscosity in mixtures. Sci Rep 14, 31857 (2024)

    2024

  16. [25]

    S., Goodman, J

    Wigh, D. S., Goodman, J. M. & Lapkin, A. A. A review of molecular representation in the age of machine learning. WIREs Comput Mol Sci 12, e1603 (2022)

    2022

  17. [26]

    & Blaschke, T

    Chen, H., Engkvist, O., Wang, Y ., Olivecrona, M. & Blaschke, T. The rise of deep learning in drug discovery. Drug Discovery Today 23, 1241–1250 (2018)

    2018

  18. [27]

    & Hinton, G

    LeCun, Y ., Bengio, Y . & Hinton, G. Deep learning. Nature 521, 436–444 (2015)

    2015

  19. [28]

    Unke, O. T. et al. Machine Learning Force Fields. Chem. Rev. 121, 10142–10186 (2021)

    2021

  20. [29]

    Specht, T. et al. HANNA: hard-constraint neural network for consistent activity coefficient prediction. Chem. Sci. 15, 19777–19786 (2024)

    2024

  21. [30]

    Keith, J. A. et al. Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems. Chem. Rev. 121, 9816–9872 (2021)

    2021

  22. [31]

    Deng, J. et al. A systematic study of key elements underlying molecular property prediction. Nat Commun 14, 6395 (2023)

    2023

  23. [32]

    & Engkvist, O

    David, L., Thakkar, A., Mercado, R. & Engkvist, O. Molecular representations in AI-driven drug discovery: a review and practical guide. J Cheminform 12, 56 (2020)

    2020

  24. [33]

    Ross, J. et al. Large-scale chemical language representations capture molecular structure and properties. Nat Mach Intell 4, 1256–1264 (2022)

    2022

  25. [34]

    & Bjerrum, E

    Irwin, R., Dimitriadis, S., He, J. & Bjerrum, E. J. Chemformer: a pre-trained transformer for computational chemistry. Mach. Learn.: Sci. Technol. 3, 015022 (2022)

    2022

  26. [35]

    Zhou, G. et al. Uni-mol: A universal 3d molecular representation learning framework. in The eleventh international conference on learning representations (2023)

    2023

  27. [36]

    Song, B. et al. A systematic review of molecular representation learning foundation models. Briefings in Bioinformatics 27, bbaf703 (2026)

    2026

  28. [37]

    Cheng, Z. et al. Automatic Screen-out of Ir(III) Complex Emitters by Combined Machine Learning and Computational Analysis. Preprint at https://doi.org/10.26434/chemrxiv-2023-q2jkb (2023)

    2023 doi

  29. [38]

    Ou, Q. et al. High-accuracy physical property prediction for pure organics via molecular representation learning: bridging data to discovery. npj Comput Mater 11, 224 (2025)

    2025

  30. [39]

    Zeng, B. et al. Uni-ELF: A Multi-level representation learning framework for electrolyte formulation design. arXiv preprint arXiv:2407.06152 (2024). Article

    2024 arXiv

  31. [40]

    RDKit: A software suite for cheminformatics, computational chemistry, and predictive modeling

    Landrum, G. RDKit: A software suite for cheminformatics, computational chemistry, and predictive modeling. (2013)

    2013

  32. [43]

    Bair, S. S. High Pressure Rheology for Quantitative Elastohydrodynamics. (Elsevier, 2019)

    2019

  33. [44]

    & Pollice, R

    Tretiakov, S., Nigam, A. & Pollice, R. Studying Noncovalent Interactions in Molecular Systems with Machine Learning. Chem. Rev. 125, 5776–5829 (2025)

    2025

  34. [45]

    R., Mogilner, I

    Ruderman, G., Caffarena, E. R., Mogilner, I. G. & Tolosa, E. J. Hydrogen Bonding of Carboxylic Acids in Aqueous Solutions—UV Spectroscopy, Viscosity, and Molecular Simulation of Acetic Acid. Journal of Solution Chemistry 27, 935–948 (1998)

    1998

  35. [46]

    Zhang, C. et al. Correlation between the Molecular Structure and Viscosity Index of CTL Base Oils Based on Ridge Regression. ACS Omega 7, 18887–18896 (2022)

    2022

  36. [47]

    G., Sun, T

    Bleazard, J. G., Sun, T. F. & Teja, A. S. The thermal conductivity and viscosity of acetic acid-water mixtures. Int J Thermophys 17, 111–125 (1996). Article Supplementary Information for: 3D molecular representation learning for organic mixtures: viscosity and density predicti...

    1996

  37. [48]

    Cibulka

    V olumetric properties of mixtures and solutions Binary liquid systems of nonelectrolytes / I. Cibulka ... Ed. by H. V . Kehiaian. in (eds Kehiaian, H. V . & Cibulka, I.) (Springer, Berlin Heidelberg New York, 2009)

    2009

  38. [49]

    Physical chemistry Viscosity of pure organic liquids and binary liquid mixtures: supplement to IV/18 / ed.: M. D. Lechner. Author: Ch. Wohlfarth. in (eds Lechner, M. D. & Wohlfarth, C.) (Springer, Berlin Heidelberg New York, 2009)

    2009

  39. [2020]

    Ind. Eng. Chem. Res. 60, 4987–5013 (2021)

    2021

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.