From pixels to phenotypes: Integrating image-based profiling with cell health data as BioMorph features improves interpretability

Molecular Biology of the Cell, Journal Year: 2024, Volume and Issue: 35(3)

Published: Jan. 3, 2024

Cell Painting assays generate morphological profiles that are versatile descriptors of biological systems and have been used to predict in vitro vivo drug effects. However, features extracted from classical software such as CellProfiler based on statistical calculations often not readily biologically interpretable. In this study, we propose a new feature space, which call BioMorph, maps these with readouts comprehensive Health assays. We validated the resulting BioMorph space effectively connected compounds only associated their bioactivity but deeper insights into phenotypic characteristics cellular processes given bioactivity. The revealed mechanism action for individual compounds, including dual-acting emetine, an inhibitor both protein synthesis DNA replication. Overall, offers relevant way interpret cell derived using hypotheses experimental validation.

Language: Английский

---

A Decade in a Systematic Review: The Evolution and Impact of Cell Painting

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: May 7, 2024

ABSTRACT High-content image-based assays have fueled significant discoveries in the life sciences past decade (2013-2023), including novel insights into disease etiology, mechanism of action, new therapeutics, and toxicology predictions. Here, we systematically review substantial methodological advancements applications Cell Painting. Advancements include improvements Painting protocol, assay adaptations for different types perturbations applications, improved methodologies feature extraction, quality control, batch effect correction. Moreover, machine learning methods recently surpassed classical approaches their ability to extract biologically useful information from images. data been used alone or combination with other - omics decipher action a compound, its toxicity profile, many biological effects. Overall, key advances expanded Painting’s capture cellular responses various perturbations. Future will likely lie advancing computational experimental techniques, developing publicly available datasets, integrating them high-content types.

Language: Английский

---

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted In Vivo and In Vitro Data

Chemical Research in Toxicology, Journal Year: 2024, Volume and Issue: 37(8), P. 1290 - 1305

Published: July 9, 2024

Drug-induced liver injury (DILI) has been a significant challenge in drug discovery, often leading to clinical trial failures and necessitating withdrawals. Over the last decade, existing suite of

Language: Английский

---

PKSmart: An Open-Source Computational Model to Predictin vivoPharmacokinetics of Small Molecules

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Feb. 7, 2024

ABSTRACT Drug exposure is a key contributor to the safety and efficacy of drugs. It can be defined using human pharmacokinetics (PK) parameters that affect blood concentration profile drug, such as steady-state volume distribution (VDss), total body clearance (CL), half-life (t½), fraction unbound in plasma (fu) mean residence time (MRT). In this work, we used molecular structural fingerprints, physicochemical properties, predicted animal PK data features model VDss, CL, t½, fu MRT for 1,283 unique compounds. First, [VDss, fu] rats, dogs, monkeys 372 compounds fingerprints properties. Second, Morgan Mordred descriptors hyperparameter-optimised Random Forest algorithm predict parameters. When validated repeated nested cross-validation, VDss was best with an R 2 0.55 Geometric Mean Fold Error (GMFE) 2.09; CL accuracies =0.31 GMFE=2.43, =0.61 GMFE=2.81, =0.28 GMFE=2.49, t½ GMFE=2.46 models combining We evaluated external test set comprising 315 (R =0.33 GMFE=2.58) =0.45 GMFE=1.98). compared our proprietary pharmacokinetic from AstraZeneca found predictions were similar Pearson correlations ranging 0.77-0.78 0.46-0.71 (dog rat) fu. To knowledge, first work publicly releases on par industry-standard models. Early assessment integration properties are crucial, DMTA cycles, which possible study based input only chemical structures. developed webhosted application PKSmart ( https://broad.io/PKSmart ) users access web browser all code also downloadable local use. Abstract Figure Figure: For TOC Only

Language: Английский

---

Evaluating feature extraction in ovarian cancer cell line co-cultures using deep neural networks

Communications Biology, Journal Year: 2025, Volume and Issue: 8(1)

Published: Feb. 25, 2025

Abstract Single-cell image analysis is crucial for studying drug effects on cellular morphology and phenotypic changes. Most studies focus single cell types, overlooking the complexity of interactions. Here, we establish an pipeline to extract features cancer cells cultured with fibroblasts. Using high-content imaging, analyze oncology library across five fibroblast line co-culture combinations, generating 61,440 images ∼170 million single-cell objects. Traditional phenotyping CellProfiler achieves average enrichment score 62.6% mechanisms action, while pre-trained neural networks (EfficientNetB0 MobileNetV2) reach 61.0% 62.0%, respectively. Variability in scores may reflect use multiple concentrations since not all induce significant morphological changes, as well genetic context treatment. Our study highlights nuanced drug-induced variations underscores heterogeneity ovarian lines their response complex environments.

Language: Английский

---

DICTrank Is a Reliable Dataset for Cardiotoxicity Prediction Using Machine Learning Methods

Chemical Research in Toxicology, Journal Year: 2025, Volume and Issue: unknown

Published: March 27, 2025

Drug-induced cardiotoxicity (DICT) is a significant challenge in drug development and public health. DICT can arise from various mechanisms; New Approach Methods (NAMs), including quantitative structure-activity relationships (QSARs), have been extensively developed to predict based solely on individual mechanisms (e.g., hERG-related cardiotoxicity) due the availability of datasets limited specific mechanisms. While these efforts significantly contributed our understanding cardiotoxicity, assessment remains challenging, suggesting that approaches focusing isolated may not provide comprehensive evaluation. To address this, we previously DICTrank, largest dataset for assessing overall liability humans FDA labels. In this study, evaluated utility DICTrank QSAR modeling using five machine learning methods─Logistic Regression (LR), K-Nearest Neighbors, Support Vector Machines, Random Forest (RF), extreme gradient boosting (XGBoost)─which vary algorithmic complexity explainability. reflect real-world scenarios, models were trained drugs approved before within 2005 risk those thereafter. observed no clear association between prediction performance model complexity, LR XGBoost achieved best results with DICTrank. Additionally, significant-feature analyses RF provided novel insights into mechanisms, revealing properties associated descriptors such as "structural topological", "polarizability", "electronegativity" DICT. Moreover, found varied by therapeutic category, need tailor accordingly. conclusion, study demonstrated robustness reliability methods.

Language: Английский

---

Improved Detection of Drug-Induced Liver Injury by Integrating Predictedin vivoandin vitroData

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 12, 2024

Drug-induced liver injury (DILI) has been significant challenge in drug discovery, often leading to clinical trial failures and necessitating withdrawals. The existing suite of vitro proxy-DILI assays is generally effective at identifying compounds with hepatotoxicity. However, there considerable interest enhancing silico prediction DILI because it allows for the evaluation large sets more quickly cost-effectively, particularly early stages projects. In this study, we aim study ML models that first predicts nine labels then uses them as features addition chemical structural predict DILI. include

Language: Английский

---

ADMET-AI Enables Interpretable Predictions of Drug-Induced Cardiotoxicity

Circulation, Journal Year: 2025, Volume and Issue: 151(3), P. 285 - 287

Published: Jan. 21, 2025

Language: Английский

---

Semisupervised Learning to Boost hERG, Nav1.5, and Cav1.2 Cardiac Ion Channel Toxicity Prediction by Mining a Large Unlabeled Small Molecule Data Set

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: 64(16), P. 6410 - 6420

Published: Aug. 7, 2024

Predicting drug toxicity is a critical aspect of ensuring patient safety during the design process. Although conventional machine learning techniques have shown some success in this field, scarcity annotated data poses significant challenge enhancing models' performance. In study, we explore potential leveraging large unlabeled small molecule sets using semisupervised to improve cardiotoxicity predictive performance across three cardiac ion channel targets: voltage-gated potassium (hERG), sodium (Nav1.5), and calcium (Cav1.2). We extensively mined ChEMBL database, comprising approximately 2 million molecules, then employed construct robust classification models for purpose. achieved boost on highly diverse (i.e., structurally dissimilar) test all targets. Using our built models, screened whole database set FDA-approved drugs, identifying several compounds with activity. To ensure broad accessibility usability both technical nontechnical users, developed cross-platform graphical user interface that allows users make predictions gain insights into drugs other molecules. The software made available as open source under permissive MIT license at https://github.com/issararab/CToxPred2.

Language: Английский

---

Cell Painting: a decade of discovery and innovation in cellular imaging

Nature Methods, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 5, 2024

Language: Английский

---