Leveraging Cell Painting Images to Expand the Applicability Domain and Actively Improve Deep Learning Quantitative Structure–Activity Relationship Models

Abstract

The search for chemical hit material is a lengthy and increasingly expensive drug discovery process. To improve it, ligand-based quantitative structure–activity relationship models have been broadly applied to optimize primary and secondary compound properties. Although these models can be deployed as early as the stage of molecule design, they have a limited applicability domainif the structures of interest differ substantially from the chemical space on which the model was trained, a reliable prediction will not be possible. Image-informed ligand-based models partly solve this shortcoming by focusing on the phenotype of a cell caused by small molecules, rather than on their structure. While this enables chemical diversity expansion, it limits the application to compounds physically available and imaged. Here, we employ an active learning approach to capitalize on both of these methods’ strengths and boost the model performance of a mitochondrial toxicity assay (Glu/Gal). Specifically, we used a phenotypic Cell Painting screen to build a chemistry-independent model and adopted the results as the main factor in selecting compounds for experimental testing. With the additional Glu/Gal annotation for selected compounds we were able to dramatically improve the chemistry-informed ligand-based model with respect to the increased recognition of compounds from a 10% broader chemical space.

Detection of a Mitochondrial Stress Phenotype using the Cell Painting Assay

“…CPA has been employed to study cell and mitochondrial toxicity (Dahlin et al, 2023;Garcia de Lomana et al, 2023;Herman et al, 2023;Seal et al, 2022;Trapotsi et al, 2022). By using CPA profiles, gene expression signatures, chemical structural information and mitochondrial toxicity data, Seal et al could distinguish between mitochondrial toxicants and non-toxicants based on the morphological profiles (Seal et al, 2022).…”

Introduction to the Special Issue: AI Meets Toxicology

“…14 In addition to models relying purely on the chemical structure, there is a notable trend of bringing in additional modalities to improve or inform predictive models. 20,21 In the following, we provide an overview of the AI approaches used in the publications contained in the SI. The methodological overview presented in this section highlights the variety of AI approaches employed in toxicology, showcasing the potential for combining different methods to enhance predictive performance and expand the scope of toxicity modeling.…”

“…The methods described in this special issue cover a wide range of AI methods ranging from expert systems, ,, over similarity measures including read-across methods, − to classical machine learning such as random forests (RF), support vector machines (SVM), and artificial neural networks (ANN) − ,, to deep learning (DL) methods ,,− , including equivariant neural networks, deep generative models, and even large language models . In addition to models relying purely on the chemical structure, there is a notable trend of bringing in additional modalities to improve or inform predictive models. , In the following, we provide an overview of the AI approaches used in the publications contained in the SI. The methodological overview presented in this section highlights the variety of AI approaches employed in toxicology, showcasing the potential for combining different methods to enhance predictive performance and expand the scope of toxicity modeling.…”

“…Most of the presented methods perform prediction tasks on molecules (for exceptions, see paragraphs below) and, thus, have to use representations of molecules. While for data handling, the simplified molecular-input line-entry system (SMILES) representation is often used, the predominant representations of molecules for modeling are still extended connectivity fingerprint (ECFP) or Morgan fingerprints. − ,− Several publications use graph neural networks that operate on the molecular graph. , Aside from the chemical structure, there is a growing tendency to incorporate biological characterizations and read-outs, for example, via cell morphology , or transcriptomics . The utilization of diverse representations, ranging from molecular structures to biological features, enhances the predictive models showcased in this section and could improve the comprehensive understanding of toxicological properties.…”

“…Garcia de Lomana et al developed machine learning models for predicting mitochondrial toxicity by incorporating curated data sets and leveraging morphological features from a cell painting screen, which improved the quality of predictions, while exploring the challenges and opportunities of using this approach. Herman et al used an active learning approach combining ligand-based quantitative structure–activity relationship models and a phenotypic cell painting screen to improve the model performance for a mitochondrial toxicity assay, enhancing chemical diversity coverage and expanding the recognition of compounds from a broader chemical space. Bussola et al have presented PathologAI, a weakly supervised deep learning framework built for classifying whole slide images in animal pathology, which was applied specifically for the prediction of necrosis in rat liver images.…”