Skip to content
DraconiticResearch
Search Sources API
197,688 studies31 sourcesUpdated 34 min ago
PreprintCC BY-NCChemRxiv

Potency Prediction via Probabilistic Modeling of Dose–Response Curves

  • Anton O’Nils1,2,3
  • Stefano Sarao Mannelli4,5,6,7
  • Ola Engkvist1
  • Jens Sjölund2,3
  • Annie M. Westerlund1
7 affiliations
  1. 1Molecular AI,, AstraZeneca Gothenburg, Pepparedsleden 1,431 83, Mölndal, Sweden
  2. 2Department of Information Technology
  3. 3Uppsala University
  4. 4Department of Computer Science and Engineering
  5. 5Chalmers University of Technology
  6. 6School of Computer Science and Applied Mathematics
  7. 7University of the Witwatersrand
Open at sourcePDFDOICite

Abstract

Predicting molecular potency from chemical structure is an important task in drug discovery, and is almost universally framed as scalar regression on the half-maximal inhibitory concentration (IC50) derived from a dose–response experiment. This formulation collapses the full experimental signal into a single number and implicitly assumes well-behaving data, even though real assay measurements might be imbalanced, noisy, and potentially censored. Because public bioactivity datasets report only the summary pIC50, the alternative of predicting the entire dose–response curve has remained essentially unexplored. Here, using an in-house dataset containing the full curves, we present a novel probabilistic approach to potency prediction that models the entire dose–response curve as the cumulative distribution function of a latent concentration variable. We benchmark it against two alternative curve-prediction models—a multi-task regression over the concentration grid and a parametric four-parameter logistic model—and against descriptor-and graph-based scalar baselines. The probabilistic model achieves the best performance on all metrics, demonstrating the competitiveness of curve-level modeling over the scalar formulation. Its advantage is two-fold. First, it exposes useful curve information, such as slope, asymptotes, and shape, that is otherwise lost in the scalar collapse. Second, it learns directly from raw measurements rather than a fitted summary statistic that itself carries uncertainty and noise. Altogether, these results show that modeling the full dose–response curve offers a robust and information-preserving alternative to conventional scalar potency prediction.

Related

  1. FlowBatt: Flow Matching for Probabilistic Battery Degradation Prediction
    Hamidreza Eivazi, Eslam Sharaawy, Stefan Wittek +3
    arXivPreprintarXiv licenceFull textPhysical chemistryarXiv
  2. Numerical Modeling of the Ultrafast Plasmonic Response of Titanium Nitride Nanostructures
    Silvia Rotta Loria, Andrea Schirato, Giulia Crotti +4
    Journal of Physical Chemistry CPeer-reviewedLicence not stated1 citedMaterials & polymersTheoretical & computational chemistryHAL
  3. PCIM-DTA: pairwise conditional interaction modeling for drug–target affinity prediction under cold-start scenarios
    Ziyou Zhou, Min Chen, 周文剑 +1
    BioinformaticsPeer-reviewedCC BYFull text8 citedMedicinal chemistry & drug discoveryOpenAlex
  4. From Reinforcement to Prediction: Mechanical Behavior and Modeling of CNT-Reinforced Cementitious Composites
    Nicolò Rimondotto, Alice Rossi, Matteo Pavese +1
    NanomaterialsPeer-reviewedCC BY 4.0Materials & polymersCrossref
  5. HierL2GDTA: Hierarchical Interaction Modeling with Local and Global Features for Drug–Target Affinity Prediction
    Kaili Wang, Yaoxiang Yin, Yanglan Gan +2
    MoleculesPeer-reviewedCC BY 4.0General chemistryCrossref

Cite

BibTeXRIS
BibTeX
@misc{onils2026potency,
  title = {{Potency Prediction via Probabilistic Modeling of Dose–Response Curves}},
  author = {Anton O’Nils and Stefano Sarao Mannelli and Ola Engkvist and Jens Sjölund and Annie M. Westerlund},
  howpublished = {ChemRxiv},
  year = {2026},
  month = sep,
  doi = {10.26434/chemrxiv.15008694/v1},
  url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008694/v1},
  note = {CC BY-NC},
}
RIS
TY  - UNPB
TI  - Potency Prediction via Probabilistic Modeling of Dose–Response Curves
AU  - Anton O’Nils
AU  - Stefano Sarao Mannelli
AU  - Ola Engkvist
AU  - Jens Sjölund
AU  - Annie M. Westerlund
PY  - 2026
DA  - 2026/09/11
PB  - ChemRxiv
DO  - 10.26434/chemrxiv.15008694/v1
UR  - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008694/v1
AB  - Predicting molecular potency from chemical structure is an important task in drug discovery, and is almost universally framed as scalar regression on the half-maximal inhibitory concentration (IC50) derived from a dose–response experiment. This formulation collapses the full experimental signal into a single number and implicitly assumes well-behaving data, even though real assay measurements might be imbalanced, noisy, and potentially censored. Because public bioactivity datasets report only the summary pIC50, the alternative of predicting the entire dose–response curve has remained essentially unexplored. Here, using an in-house dataset containing the full curves, we present a novel probabilistic approach to potency prediction that models the entire dose–response curve as the cumulative distribution function of a latent concentration variable. We benchmark it against two alternative curve-prediction models—a multi-task regression over the concentration grid and a parametric four-parameter logistic model—and against descriptor-and graph-based scalar baselines. The probabilistic model achieves the best performance on all metrics, demonstrating the competitiveness of curve-level modeling over the scalar formulation. Its advantage is two-fold. First, it exposes useful curve information, such as slope, asymptotes, and shape, that is otherwise lost in the scalar collapse. Second, it learns directly from raw measurements rather than a fitted summary statistic that itself carries uncertainty and noise. Altogether, these results show that modeling the full dose–response curve offers a robust and information-preserving alternative to conventional scalar potency prediction.
ER  -