Cell Painting and machine learning for mechanistic profiling of hepatotoxic botanicals and their constituents

Ollitrault, G.; Mitchell, C. A.; Ferguson, S. S.; Roe, A. L.; White, B.; Titterton, K.; Ewald, J.

doi:10.64898/2026.10.05.756246

Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.10.05.756246

Cell Painting and machine learning for mechanistic profiling of hepatotoxic botanicals and their constituents

Ollitrault, G.·Mitchell, C. A.·Ferguson, S. S.·Roe, A. L.·White, B.·Titterton, K.·Ewald, J.

View original source ↗Export metadata ↓

Abstract

BACKGROUND Botanicals are widely consumed as herbal remedies and plant-derived products, but their complex and variable chemical composition presents major challenges for safety assessment. Although hepatotoxicity has been reported for several botanicals, the constituent compounds responsible and the underlying mechanisms often remain poorly characterised. OBJECTIVES We aimed to determine whether Cell Painting could identify candidate constituent drivers of botanical bioactivity and generate hypotheses about potential mechanisms of action in primary human hepatocytes. METHODS We profiled 13 botanical extracts and 85 known constituents across multiple concentrations using Cell Painting alongside assays of metabolic activity and membrane integrity. To interpret morphological phenotypes, we trained machine-learning models using a Cell Painting reference atlas of >1,000 chemicals linked to ToxCast/Tox21 molecular and cellular bioactivity annotations. RESULTS Cell Painting detected bioactivity in 11/13 botanical extracts and 46/85 constituents, compared with 9/13 extracts and 40/85 constituents in the cell-health assays, and generally detected activity at lower concentrations. For 9/13 extracts, at least one constituent was active at a lower concentration than the corresponding botanical extract, supporting the contribution of a subset of potent constituents to mixture-level responses. Aristolochia and goldenseal were the most potent extracts across Cell Painting and cell-health assays and contained highly active constituents. Reference-atlas models generated specific mechanistic hypotheses, including cytotoxicity-, endocrine-, and genotoxicity-associated responses for several botanicals and constituents. CONCLUSION These results support Cell Painting combined with machine learning as a sensitive approach for characterising cellular responses to botanicals, prioritising candidate constituent drivers of hepatotoxicity-associated phenotypes, and generating testable hypotheses about potential mechanisms of action.

Explore related subjects

pharmacology and toxicology Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Ollitrault, G., Mitchell, C. A., Ferguson, S. S., Roe, A. L., White, B., Titterton, K., Ewald, J.. 2026-10-12. Cell Painting and machine learning for mechanistic profiling of hepatotoxic botanicals and their constituents. https://doi.org/10.64898/2026.10.05.756246

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

A mechanistic classification of GNAO1 mutations links Gβ γ sequestration to hyperkinetic disease and treatment response

Go is the most prevalent G protein subunit in the brain and is critical for signal transduction through G protein-coupled receptors. Mutations in the gene encoding Go (GNAO1) cause a rare neurodevelopmental disorder with involuntary movements (NEDIM) and, in some patients, developmental and epileptic encephalopathy (DEE). Here, we integrate human genetic analysis with cellular assays to define mechanistic classes of GNAO1 mutations and in vivo approaches to evaluate therapeutic strategies. We grouped variants into two functional categories: those predicted to sequester G{beta}{gamma} and those with reduced Go expression and therefore increased G{beta}{gamma} signaling. Analysis of patient data revealed that mutations predicted to impair G{beta}{gamma} release are significantly associated with NEDIM. Using the GNAO1R209H/+ mouse as a model of NEDIM, we evaluated two small molecules: caffeine, a nonselective adenosine receptor antagonist, and gallein, a selective modulator of G{beta}{gamma}-effector interactions. Both compounds significantly reduced hyperlocomotion in the open field test. Whole-cell patch-clamp electrophysiology revealed sex- and cell type-specific hyperexcitability in spiny projection neurons within the dorsolateral striatum of GNAO1R209H/+ mice. Together, these findings implicate dysfunction of dorsolateral striatal pathways in GNAO1-related disorder and suggest that mechanistic understanding of disease variants can guide targeted therapeutic strategies.

pharmacology and toxicology↗

bioRxiv2026

Screening and evaluation of medicine food homology herbal combinations for ameliorating chemotherapy induced ovarian aging under an integrated multi-strategy framework

Abstract Background: Ovarian aging poses a serious threat to women's reproductive health. Medicine food homology (MFH) herbs, owing to their favorable safety profile and suitability for long-term use, have become important interventional resources. However, conventional formula screening relies largely on empirical statistics and lacks mechanistic evidence and a systematic screening strategy. Methods: In this study, we integrated single cell transcriptomic analysis, human use evidence mining, UPLC/Q TOF MS chemical profiling and pretrained protein molecular representation models to establish a multi strategy framework for the discovery of MFH herbal combinations. Single cell transcriptomic datasets related to ovarian aging from the GEO database were analyzed to identify key pathological processes and candidate targets. High-frequency MFH herbs were selected by mining classical prescriptions and clinical literature. By comparing six protein molecular pretrained representation combinations, ESM T5 was identified for compound target affinity prediction and validated by molecular dynamics simulation and surface plasmon resonance (SPR) assays, ultimately identifying the Astragalus ginseng (Huangqi Renshen) herb pair (HXF) as the core candidate. Functional expression validation of key targets was performed, and the efficacy of HXF and the expression of related molecules were verified in a D gal induced KGN granulosa cell senescence model and a cyclophosphamide-induced rat model of diminished ovarian function. Results: Single cell analysis revealed that ovarian aging was characterized by impaired steroidogenesis, enhanced MDK NCL communication between granulosa cells and immune cells, and extracellular matrix remodeling, with CYP19A1, SERPINE1 and MDK as key targets. By further integrating human-use evidence mining and UPLC/Q TOF MS compound annotation, an MFH compound library comprising 712 molecules was established. ESM T5 prediction, molecular dynamics (MD) simulation and SPR assays confirmed that multiple constituents of Astragalus and ginseng exhibited favorable binding activity toward key targets. At the cellular level, knockdown of MDK or SERPINE1 attenuated the senescence phenotype, whereas knockdown of CYP19A1 aggravated it. In vitro and in vivo experiments demonstrated that Astragalus ginseng ameliorated granulosa cell senescence and diminished ovarian function in rats, restored CYP19A1 related steroidogenesis, downregulated inflammation- and matrix remodeling related molecules such as SERPINE1 and IL18, and reduced MDK NCL colocalization signals. Conclusion: This study established a multi-strategy screening framework integrating disease mechanism decoding, human use evidence constraints and pretrained model prediction, and identified Astragalus ginseng as a candidate herb pair for ameliorating ovarian aging. Its effects may be associated with promoting steroidogenesis, suppressing SERPINE1/IL18 related inflammation and matrix remodeling, and modulating MDK NCL signaling. These findings provide a strategic reference for the mechanism oriented discovery of MFH herbal combinations.

pharmacology and toxicology↗

bioRxiv2026

BRET-Based Ligand Binding at Recombinant and Endogenous Dopamine D2 Receptors

Dopamine D2 receptor (D2R) regulates diverse biological processes and is an important therapeutic target, underscoring the need for robust methods to quantify ligand binding at this receptor. Bioluminescence resonance energy transfer (BRET)-based ligand binding assays have emerged as a powerful platform for quantifying binding in living cells, but their application to D2R has been limited by the availability of suitable fluorescent tracers. Here, we developed three far-red fluorescent antagonists (CELT-240, CELT-383, and CELT-385) as BRET tracers for D2R. All three compounds produced specific, saturable D2R binding with nanomolar affinity. Further characterization of CELT-385 demonstrated its utility for kinetic and competition binding measurements and revealed high affinity binding at the closely related dopamine D3 receptor (D3R). We also extended BRET-based ligand binding beyond genetically-modified receptors using bystander BRET, enabling measurements at unmodified recombinant and endogenously expressed D2R. Together, these findings provide new tools for quantitative D2R/D3R ligand binding and a template for extending BRET-based ligand binding to endogenously expressed GPCRs.

pharmacology and toxicology↗