Pith. sign in

REVIEW 5 major objections 5 minor 37 references

A single multi-task model with quantum-informed features and learnable per-task loss weights outperforms single-task baselines on 12 of 13 ADMET classification endpoints.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-05 05:56 UTC pith:6LSCQIOB

load-bearing objection A useful 13-task ADMET MTL benchmark, but the learnable weighting mechanism as written makes the headline results numerically implausible; needs major clarification or code release before I'd trust it. the 5 major comments →

arxiv 2509.04601 v1 pith:6LSCQIOB submitted 2025-09-04 cs.LG cs.AI

Quantum-Enhanced Multi-Task Learning with Learnable Weighting for Pharmacokinetic and Toxicity Prediction

Han Zhang , Fengji Ma , Jiamin Su , Xinyue Yang , Lei Wang , Wen-Cai Ye , Li Liu This is my paper
classification cs.LG cs.AI
keywords multi-task learningADMET predictionquantum chemical descriptorsGFN2-xTBlearnable task weightingD-MPNNdrug screeningTherapeutics Data Commons
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper sets out to show that one shared model can handle all 13 classification endpoints in the TDC ADMET benchmark better than training 13 models separately. The proposed QW-MTL framework takes Chemprop's directed message-passing neural network, enriches its molecular fingerprint with 200 RDKit descriptors and four GFN2-xTB quantum descriptors, and balances training with weights of the form w_t = r_t^{softplus(log β_t)}, where r_t is the task's share of labels in the batch and β_t is learned. On leaderboard-style splits it beats the single-task Chemprop-RDKit baseline on 12 of 13 tasks, with the largest gains on small datasets such as the CYP substrate and DILI endpoints, and it cuts inference time about tenfold. The paper's point is that joint training plus physically grounded electronic-structure features plus a data-scale-aware weighting rule is an effective and cheap strategy for ADMET screening.

Core claim

On the paper's own terms, the discovery is that the standard practice of building a separate predictor for each ADMET property is both more expensive and less accurate than one shared model. QW-MTL constructs a 13-task classifier whose single shared encoder is a D-MPNN; its input representation concatenates the learned molecular fingerprint, 200 RDKit-derived physicochemical descriptors, and four GFN2-xTB quantum features — dipole moment, HOMO-LUMO gap, electron count, and total energy — plus a binary mask for failed quantum computations. Task heads share the encoder, and the total loss weights each task by r_t^{softplus(log β_t)}, a batch sample-proportion raised to a learnable exponent. On

What carries the argument

The mechanism that carries the argument is the weighting rule w_t = r_t^{softplus(log β_t)}: r_t is the fraction of valid labels in a batch belonging to task t, and β_t is a per-task parameter learned jointly with the network. Softplus keeps the exponent positive, so the model can flatten or steepen each task's scale-based weight automatically. The second carrier is the fused descriptor vector — D-MPNN fingerprint plus 200 RDKit descriptors plus four quantum descriptors plus a 4-bit missing mask — which gives all 13 task heads simultaneous access to topology, global physicochemical properties, and electronic structure.

Load-bearing premise

The load-bearing premise is that the particular learnable weighting formula — task sample fraction raised to a learned exponent — is what produces the reported gains, and the paper only tests it against equal weighting, not against other standard task-balancing methods.

What would settle it

Run the identical 13-task protocol with the learnable exponent β replaced by a fixed inverse-scale weight (w_t = 1/n_t) or by an uncertainty-based weight; if average performance equals QW-MTL, the learnable exponent is not the cause. Separately, run QW-MTL with all four quantum descriptors forced to zero while keeping the mask active; if performance is unchanged, the quantum features are not the cause.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If QW-MTL is right, a drug-screening pipeline can replace 13 single-task models with one 384k-parameter model, lowering inference cost by roughly 10.5x on a 10,000-molecule batch.
  • Small-data endpoints are the main beneficiaries: CYP2C9-S, CYP2D6-S, and DILI each improve about 7% over the single-task baseline, showing that multi-task sharing transfers supervision from larger tasks.
  • The learned exponents track dataset size closely (r = 0.950), so no manual loss-balancing schedule is needed; the model self-tunes its task weights during training.
  • The full combination (multi-task plus quantum descriptors plus learnable β) helps 10 of 13 tasks over the plain multi-task RDKit model, implying the two additions are complementary rather than redundant.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves open whether the specific exponential form matters: replacing w_t with a fixed 1/n_t prior, or with uncertainty-based or gradient-normalized weights, would isolate the causal role of the learnable exponent.
  • Because roughly 10% of molecules lack valid quantum descriptors, a masking experiment that zeros all four QC features during inference could reveal how much of the gain is genuinely electronic-structure information versus architectural capacity.
  • Since β_t correlates with sample size, a plausible simplification is a closed-form β(n_t) fitted once, potentially removing the extra parameters without losing the balancing effect.
  • Extending the same joint-training protocol to the regression ADMET endpoints (the excluded 'E' tasks) is a natural test of whether the shared encoder generalizes to a full ADMET profile.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes QW-MTL, a unified multi-task learning framework for 13 TDC ADMET classification tasks. It builds on a Chemprop D-MPNN with RDKit and GFN2-xTB quantum descriptors, and introduces a learnable sample-proportion-based loss weighting w_t = r_t^{softplus(log β_t)}. The authors report gains over a single-task Chemprop-RDKit baseline on 12 of 13 tasks, top leaderboard ranks on several tasks, a 10.5x inference speedup, and ablations separating the contributions of multi-task training, quantum descriptors, and learnable weighting.

Significance. If confirmed, the work would provide a useful standardized multi-task evaluation on all 13 TDC ADMET classification tasks using official leaderboard splits, and the combination of quantum-informed features with MTL is of interest. The efficiency measurements (10.5x, 14x, 12x speedups in Appendix B) are concrete and reproducible in principle. However, the manuscript is not internally consistent: Eq. (2) assigns tiny weights to the exact low-resource tasks whose gains are emphasized; the leaderboard tables in the main text and appendix contradict each other; the appendix table supporting Figure 2 is misaligned; and the term 'significantly' is not supported by the reported error bars for most tasks. The core idea may be salvageable, but the evidence as written is not reliable.

major comments (5)
  1. [Methodology, Eq. (2); Table 5] Equation (2) defines w_t = r_t^{softplus(log β_t)}. Using the data scales in Table 5 (total ≈ 52,329), DILI has r ≈ 0.0091 and β = 3.0, so softplus(log 3) = log 4 ≈ 1.386 and w_DILI ≈ 1.5×10⁻³. For CYP2C9-S (r ≈ 0.0127, β = 6.0), softplus(log 6) = log 7 ≈ 1.946 and w ≈ 2.0×10⁻⁴. These weights are orders of magnitude below those of large tasks (e.g., CYP2D6-I w ≈ 0.146, Ames w ≈ 0.176). The statement in 'Sample-Aware Task Weighting Analysis' that smaller β 'ensures contributions remain non-negligible' is contradicted by these numbers. With such small weights, the DILI head receives negligible gradient from its own loss, so the reported DILI improvement (0.873→0.932 in Table 1; 0.928→0.939 for +Learnable-β in Table 4) is difficult to explain by the stated mechanism. Please report the actual weight vectors, state whether weights are renormalized, and clarify whether Eq. (2) is mis-specified
  2. [Experiments, Table 1; Abstract] The abstract claims QW-MTL 'significantly outperforms' the single-task baseline on 12/13 tasks, but the reported ±1 SD intervals overlap for all tasks except DILI. Examples: Pgp 0.895±0.013 vs 0.913±0.013; BBB 0.905±0.008 vs 0.909±0.002; CYP2D6-S 0.675±0.034 vs 0.722±0.020; CYP2C9-S 0.415±0.019 vs 0.445±0.023. Ames is a tie (0.840 vs 0.839). No significance test is reported across the repeated folds. The 'significant' wording should be softened to 'numerically higher on most tasks' or supported by paired statistical tests.
  3. [Appendix A, Table 1 vs. Table 2] The leaderboard claims are internally inconsistent. Table 2 reports DILI SOTA = ZairaChem 0.925±0.005 and QW-MTL = 0.932±0.014; Appendix A Table 1 reports DILI SOTA = MolBERT 0.878±0.007 and QW-MTL = 0.890±0.007. These cannot both be correct for the same TDC leaderboard. Appendix A also lists QW-MTL DILI as 0.890, in conflict with the main-text value of 0.932. The 'top-1 on 3 tasks, top-2 on 5 tasks' claim is therefore not currently verifiable. The authors should reconcile the tables and specify the leaderboard snapshot and version.
  4. [Appendix F, Table 4] Appendix F Table 4, presented as the numerical values for Figure 2, does not match the main text. It lists HIA as 0.675/0.705, whereas Table 1 gives the HIA baseline as 0.981 and QW-MTL as 0.989; the values 0.675/0.705 correspond to Bioavailability ma in Table 1. The rows in Appendix F Table 4 appear shifted relative to Table 1 (e.g., 'CYP2C9 Substrate' 0.873/0.928 matches DILI in Table 1). This makes Figure 2 unsupported by the supplied data and suggests a copy/paste or row-alignment error. Please regenerate the table with correct task-value alignment.
  5. [Ablation Study, Table 4; Eq. (2)] The learnable weighting component is only compared against a uniform-weight Multi-RDKit configuration. To establish that the exponential form in Eq. (2) is responsible for the gains, the authors should include established task-balancing baselines: uncertainty weighting (Kendall et al.), GradNorm, and simple inverse-frequency weights such as 1/n_t or 1/√n_t. Without this, the specific contribution of the learnable exponent is not identifiable.
minor comments (5)
  1. [Throughout] Typographical issues: 'Firstly, different' has a missing space; Table 4 contains artifacts such as '0 .989'; some inline math is malformed.
  2. [Sample-Aware Task Weighting Analysis] The text refers to 'the global weighting parameter β' but no global β is defined earlier. Please clarify the notation: β_t is per-task, and the global β is not used in Eq. (2).
  3. [Quantum Feature Integration] The 4-dimensional binary mask for missing quantum descriptors is described, but it is not reported how many molecules per task have missing descriptors, nor how the mask is treated during inference. Please provide this information.
  4. [Appendix C] The input dimension 508 is said to be 208 external descriptors + 300 learned embedding. Please spell out the composition of the 208 external descriptors (e.g., RDKit 200 + 4 quantum + 4 mask) so the count is verifiable.
  5. [References] The 'MTGL-ADMET' reference is incomplete and not formatted consistently with the rest of the bibliography.

Circularity Check

0 steps flagged

No significant circularity: predictions are evaluated on external TDC test splits; the learnable weighting is an unproven ansatz, not a self-referential derivation.

full rationale

The paper's central claims are evaluated on official TDC leaderboard test splits, which are external to the training procedure. The learnable exponents β_t in Eq. (2) are fitted parameters, but the reported AUROC/AUPRC numbers are measured on held-out test sets, so there is no fitted-input-called-prediction structure. The weighting formula w_t = r_t^{softplus(log β_t)} is an unproven ansatz (Eq. 2), not derived from first principles; an ansatz is not circular unless it is justified by a self-citation chain or reduces to the target claim. The ablation (Table 4) compares variants against a common backbone, and the 10.5× speedup is a direct runtime comparison, not a derivation from the model equations. The beta–sample-size correlation (Figure 3 / Appendix F Table 5) is descriptive post-hoc analysis, not a prediction. No self-citations are load-bearing; the cited backbones (Chemprop, ADMET-AI, GFN2-xTB, TDC) are external prior work. The internal inconsistency noted by skeptics—tiny weights for DILI/CYP2C9-S by Eq. (2) versus claimed gains—is a correctness/verifiability concern, not circularity, because it does not amount to an equation reducing to its inputs. Therefore no circular step is identified.

Axiom & Free-Parameter Ledger

1 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical entities. The main added machinery is a set of 13 learnable weighting exponents and the choice of four quantum descriptors, both fitted or chosen by hand. The transfer assumption across ADMET tasks and the ad hoc weighting form are the principal unproven premises.

free parameters (1)
  • per-task weighting exponent beta_t (through softplus(log beta_t)) = 13 values; examples: DILI 3.000, Ames 1.412, CYP2C9 Inhibition 5.892 (Appendix F Table 5)
    Learned via backpropagation on the weighted multi-task loss; determines task loss weights w_t = r_t^{softplus(log beta_t)}. These are the core mechanism of the proposed weighting scheme and are fit to the training data.
axioms (5)
  • domain assumption GFN2-xTB semi-empirical tight-binding method yields physically meaningful descriptors (dipole moment, HOMO-LUMO gap, electron count, total energy) that improve ADMET classification
    The claimed benefit of quantum descriptors rests on this toolchain's accuracy being sufficient. Invoked in 'Quantum Feature Integration'.
  • domain assumption Shared D-MPNN encoder representations transfer useful information across the 13 ADMET tasks
    The entire MTL premise; if tasks share no usable inductive bias, joint training would not help. Invoked in 'Overview of QW-MTL Framework' and the ablation in Figure 2.
  • domain assumption TDC leaderboard scaffold splits remain valid when rows are duplicated per task and merged for multi-task training
    The evaluation protocol relies on the official splits being free of leakage; stated in 'Experimental Setup'.
  • ad hoc to paper The weighting form w_t = r_t^{softplus(log beta_t)} is a suitable inductive bias
    Equation (2). The functional form is introduced without derivation from first principles; its validity is only tested empirically on these 13 tasks.
  • standard math Standard neural network training assumptions (stochastic gradient descent, backpropagation, etc.)
    Implicit in all experiments and not stated in the paper.

pith-pipeline@v1.4.0-alltime-deepseek-medium · 12422 in / 15544 out tokens · 138583 ms · 2026-08-05T05:56:18.459072+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Quantum-Enhanced Multi-Task Learning with Learnable Weighting for Pharmacokinetic and Toxicity Prediction." pith.science (2026). https://pith.science/paper/6LSCQIOB

@misc{pith2026250904601,
  author       = {Pith},
  title        = {Pith review of: Quantum-Enhanced Multi-Task Learning with Learnable Weighting for Pharmacokinetic and Toxicity Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6LSCQIOB}},
  note         = {Machine review of arXiv:2509.04601}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Prediction for ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) plays a crucial role in drug discovery and development, accelerating the screening and optimization of new drugs. Existing methods primarily rely on single-task learning (STL), which often fails to fully exploit the complementarities between tasks. Besides, it requires more computational resources while training and inference of each task independently. To address these issues, we propose a new unified Quantum-enhanced and task-Weighted Multi-Task Learning (QW-MTL) framework, specifically designed for ADMET classification tasks. Built upon the Chemprop-RDKit backbone, QW-MTL adopts quantum chemical descriptors to enrich molecular representations with additional information about the electronic structure and interactions. Meanwhile, it introduces a novel exponential task weighting scheme that combines dataset-scale priors with learnable parameters to achieve dynamic loss balancing across tasks. To the best of our knowledge, this is the first work to systematically conduct joint multi-task training across all 13 Therapeutics Data Commons (TDC) classification benchmarks, using leaderboard-style data splits to ensure a standardized and realistic evaluation setting. Extensive experimental results show that QW-MTL significantly outperforms single-task baselines on 12 out of 13 tasks, achieving high predictive performance with minimal model complexity and fast inference, demonstrating the effectiveness and efficiency of multi-task molecular learning enhanced by quantum-informed features and adaptive task weighting.

Figures

Figures reproduced from arXiv: 2509.04601 by Fengji Ma, Han Zhang, Jiamin Su, Lei Wang, Li Liu, Wen-Cai Ye, Xinyue Yang.

Figure 1
Figure 1. Figure 1: The framework of QW-MTL: All tasks share a common encoder (D-MPNN), and the learned molecular fingerprints [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance gain of multi-task over single-task. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of Single-RDKit and Multi-RDKit molecular representations using PCA. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Correlation between task sample size and the [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: PCA visualizations of molecular representations under different model settings. Each subfigure corresponds to one [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: t-SNE visualizations of molecular representations under different model settings. Each subfigure corresponds to one [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

37 extracted references · 33 canonical work pages · 2 internal anchors

  1. [1]

    Bannwarth, C.; Ehlert, S.; and Grimme, S. 2019 a . A generally applicable atomic-charge dependent London dispersion correction. Journal of Chemical Theory and Computation, 15(3): 1652--1671

    work page 2019

  2. [2]

    Bannwarth, C.; Ehlert, S.; and Grimme, S. 2019 b . GFN2-xTB—An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions. Journal of Chemical Theory and Computation, 15(3): 1652--1671

    work page 2019

  3. [3]

    Becke, A. D. 1993. Density-functional thermochemistry. III. The role of exact exchange. The Journal of Chemical Physics, 98(7): 5648--5652

    work page 1993

  4. [4]

    bin Javaid, M.; Gervens, T.; Mitsos, A.; Grohe, M.; and Rittig, J. G. 2025. Exploring data augmentation: Multi-task methods for molecular property prediction. Computers & Chemical Engineering, 201: 109253

    work page 2025

  5. [5]

    Borkotoky, S.; Joshi, A.; Kaushik, V.; and Jha, A. N. 2022. Machine learning and artificial intelligence in therapeutics and drug development life cycle. In Drug Development Life Cycle. IntechOpen

    work page 2022

  6. [6]

    Chen, J.; Chung, Y.; Tynan, J.; Cheng, C.; Yang, S.; and Cheng, A. 2024. Performance Insights for Small Molecule Drug Discovery Models: Data Scaling, Multitasking, and Generalization. arXiv preprint. To appear or preprint; check for updated venue if applicable

    work page 2024

  7. [7]

    Chen, Z.; Badrinarayanan, V.; Lee, C.-Y.; and Rabinovich, A. 2018. GradNorm: Gradient normalization for adaptive loss balancing in deep multitask networks. In Proceedings of the 35th International Conference on Machine Learning (ICML), 794--803. PMLR

    work page 2018

  8. [8]

    Daoui, O.; Elkhattabi, S.; Chtita, S.; Elkhalabi, R.; Zgou, H.; and Benjelloun, A. T. 2021. QSAR, molecular docking and ADMET properties in silico studies of novel 4,5,6,7-tetrahydrobenzo[D]-thiazol-2-Yl derivatives derived from dimedone as potent anti-tumor agents through inhibition of C-Met receptor tyrosine kinase. Heliyon, 7(7): e07463

    work page 2021

  9. [9]

    E.; Herbst, M

    Fisher, K. E.; Herbst, M. F.; and Marzouk, Y. M. 2024. Multitask methods for predicting molecular properties from heterogeneous data. The Journal of Chemical Physics, 161(1)

    work page 2024

  10. [10]

    J.; Trucks, G.; Schlegel, H.; Scuseria, G.; Robb, M.; Cheeseman, J.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G.; et al

    Frisch, M. J.; Trucks, G.; Schlegel, H.; Scuseria, G.; Robb, M.; Cheeseman, J.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G.; et al. 2009. Gaussian 09 R evision D .01. Gaussian Inc. Wallingford CT, 200

    work page 2009

  11. [11]

    S.; Riley, P

    Gilmer, J.; Schoenholz, S. S.; Riley, P. F.; Vinyals, O.; and Dahl, G. E. 2017. Neural message passing for quantum chemistry. In Proceedings of the 34th International Conference on Machine Learning (ICML), 1263--1272. PMLR

    work page 2017

  12. [12]

    Han, H.; Shaker, B.; Lee, J.; Choi, S.; Yoon, S.; Singh, M.; Basith, S.; Cui, M.; Ahn, S.; An, J.; Kang, S.; Yeom, M.; and Choi, S. 2025. Employing automated machine learning (automl) methods to facilitate the in silico admet properties prediction. Journal of Chemical Information and Modeling, 65: 3215--3225

    work page 2025

  13. [13]

    P.; Chung, Y.; Li, S.-C.; Graff, D

    Heid, E.; Greenman, K. P.; Chung, Y.; Li, S.-C.; Graff, D. E.; Vermeire, F. H.; Wu, H.; Green, W. H.; and McGill, C. J. 2024. Chemprop: A Machine Learning Package for Chemical Property Prediction. Journal of Chemical Information and Modeling, 64(1): 9--17

    work page 2024

  14. [14]

    W.; Xiao, C.; and Sun, J

    Huang, K.; Fu, T.; Gao, W.; Zhao, Y.; Roohani, Y.; Leskovec, J.; Coley, C. W.; Xiao, C.; and Sun, J. 2021. Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development. arXiv preprint arXiv:2102.09548

    Pith/arXiv arXiv 2021

  15. [15]

    Kendall, A.; Gal, Y.; and Cipolla, R. 2018. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 7482--7491. IEEE

    work page 2018

  16. [16]

    Kim, J.; Chang, W.; Ji, H.; and Joung, I. 2024. Quantum-Informed Molecular Representation Learning Enhancing ADMET Property Prediction. Journal of Chemical Information and Modeling, 64(13): 5028--5040

    work page 2024

  17. [17]

    Landrum, G. 2013. RDKit: Open-source cheminformatics. Https://www.rdkit.org

    work page 2013

  18. [18]

    Landrum, G. 2023. rdkit/rdkit: 2023\_03\_3 (Q1 2023) Release. https://zenodo.org/records/8254217. Accessed: 2023-04-01

    work page arXiv 2023

  19. [19]

    Li, P.; Hua, L.; Ma, Z.; Hu, W.; Liu, Y.; and Zhu, J. 2024. Conformalized Graph Learning for Molecular ADMET Property Prediction and Reliable Uncertainty Quantification. Journal of Chemical Information and Modeling, 64(23): 8705--8717

    work page 2024

  20. [20]

    Li, X.; and Fourches, D. 2020. Inductive transfer learning for molecular activity prediction: Next-Gen QSAR Models with MolPMoFiT. Journal of Cheminformatics, 12: 27

    work page 2020

  21. [21]

    Liang, S.; and Zhang, Y. 2020. A simple general approach to balance task difficulty in multi-task learning. arXiv preprint arXiv:2002.04792

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  22. [22]

    H.; and Lu, A

    Lin, J. H.; and Lu, A. Y. 1997. Role of pharmacokinetics and metabolism in drug discovery and development. Pharmacological Reviews, 49(4): 403--449. Comparative Study, Review

    work page 1997

  23. [23]

    Link, J.; et al. 2023. Multi-Task Learning for Graph-Based ADMET Property Prediction (MTGL-ADMET). In names, E., ed., Advances in Computational Chemistry, Volume 3, chapter 6. Springer

    work page 2023

  24. [24]

    Liu, S.; Qu, M.; Zhang, Z.; Cai, H.; and Tang, J. 2022. Structured multi-task learning for molecular property prediction. In International conference on artificial intelligence and statistics, 8906--8920. PMLR

    work page 2022

  25. [25]

    Lu, S.; Gao, Z.; He, D.; and et al. 2024. Data-driven quantum chemical property prediction leveraging 3D conformations with Uni-Mol+. Nature Communications, 15: 7104

    work page 2024

  26. [26]

    Méndez-Lucio, O.; Nicolaou, C.; and Earnshaw, B. 2022. MoIE: a molecular foundation model for drug discovery. arXiv preprint arXiv:2211.02657

    Pith/arXiv arXiv 2022

  27. [27]

    A.; Reutlinger, M.; Brandl, P.; Wang, W.; Hert, J.; and Desai, P

    Napoli, J. A.; Reutlinger, M.; Brandl, P.; Wang, W.; Hert, J.; and Desai, P. 2025. Multitask Deep Learning Models of Combined Industrial Absorption, Distribution, Metabolism, and Excretion Datasets to Improve Generalization. Molecular Pharmaceutics, 22(4): 1892--1900

    work page 2025

  28. [28]

    T.; Jeong, H.; Yang, E.; and Hwang, S

    Nguyen, A. T.; Jeong, H.; Yang, E.; and Hwang, S. J. 2021. Clinical risk prediction with temporal probabilistic asymmetric multi-task learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 9081--9091

    work page 2021

  29. [29]

    H.; and Wood, M

    Notwell, J. H.; and Wood, M. W. 2023. Admet property prediction through combinations of molecular fingerprints. arXiv preprint arXiv:2310.00174

    Pith/arXiv arXiv 2023

  30. [30]

    Rogers, D.; and Hahn, M. 2010. Extended-connectivity fingerprints. Journal of Chemical Information and Modeling, 50(5): 742--754

    work page 2010

  31. [31]

    C.; Shivnaraine, R

    Swanson, K.; Walther, P.; Leitz, J.; Mukherjee, S.; Wu, J. C.; Shivnaraine, R. V.; and Zou, J. 2024. ADMET-AI: a machine learning ADMET platform for evaluation of large-scale chemical libraries. Bioinformatics, 40(7): btae416

    work page 2024

  32. [32]

    Tang, W.; Li, M.; Zhan, Y.; and Chen, B. 2025. Adaptively multi-modal contrastive fusion network for molecular properties prediction. Engineering Applications of Artificial Intelligence, 152: 110782

    work page 2025

  33. [33]

    Tian, H.; Ketkar, R.; and Tao, P. 2022. Accurate ADMET prediction with XGBoost. arXiv preprint arXiv:2204.07532

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  34. [34]

    Waterbeemd, H.; and Gifford, E. 2003. Admet in silico modelling: towards prediction paradise? Nature Reviews Cancer, 2: 192--204

    work page 2003

  35. [35]

    Xu, S.; Xie, L.; Dai, R.; and Lyu, Z. 2025. Dumpling GNN: Hybrid GNN Enables Better ADC Payload Activity Prediction Based on the Chemical Structure. International Journal of Molecular Sciences, 26(10): 4859

    work page 2025

  36. [36]

    Yang, K.; Swanson, K.; Jin, W.; Coley, C.; Eiden, P.; Gao, H.; Guzman-Perez, A.; Hopper, T.; Kelley, B.; Mathea, M.; Palmer, A.; Settels, V.; Jaakkola, T.; Jensen, K.; and Barzilay, R. 2019. Analyzing learned molecular representations for property prediction. Journal of Chemical Information and Modeling, 59(8): 3370--3388

    work page 2019

  37. [37]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...