Research team
Expertise
- (Biomedical) image analysis, machine learning and deep learning on 2D and 3D images - High-throughput image analysis - Fluorescence microscopy - Developing methods for the phenotyping of 2D and 3D cell cultures (e.g., organoids)
High-content in-toto organoid profiling with single-cell resolution using deep learning-enhanced analysis.
Abstract
Despite technological improvements, drug discovery programs have become less successful and more expensive over time. This can in part be attributed to the rigid implementation of sub-optimal preclinical screening platforms that mainly use simple cell cultures, and toxicity and pharmacokinetics experiments with animal models. Organoids are the promise of next-generation model systems for preclinical research. The main roadblocks for organoid adoption are their lack of reproducibility and the absence of technology to characterise them in depth. We believe that robust and reproducible organoid production and analysis can only be guaranteed when organoids are characterized in toto with cellular resolution. With this project, we intend to develop a pipeline for fast cellular phenotyping of intact organoids and prepare for launching a spin-off company that offers this as a service platform to the pharma and biotech industry.Researcher(s)
- Promoter: De Vos Winnok
- Co-promoter: Sijbers Jan
- Fellow: Van De Looverbosch Tim
Research team(s)
Project type(s)
- Research Project
INFERENCE. Scalable screening platform for predicting the mode-of-action of gene perturbations based on Integrated Functional Enrichment analysis of gene expREssion aNd CEll phenotypic readouts.
Abstract
Within the classical drug discovery pipeline, early target selection and compound validation are based on simple readouts from technologies that average across large populations of cells. This strategy negates much of the total information content in the biological sample at hand, causing selection bias and attrition of promising leads. High-content microscopy holds large potential for refined mode-of-action (MoA) analysis of pharmaco-genomic perturbations. An especially information-rich readout can be obtained with Cell Painting (CP), a pipeline that is implemented in our lab and consists of automated microscopy and morphological analysis of cells stained with inexpensive fluorescent dyes. The resulting cell phenotypic signatures can be used to predict the MoA of compound treatments with high fidelity. However, by design, predictions are limited to known MoA encountered in the dataset. Furthermore, confounding factors, such as experimental noise and intercellular heterogeneity may obscure relevant biological properties. Hence, we envision a more comprehensive MoA documentation by adding a complementary information layer based on transcriptomics of the same cell culture at hand. To this end, we have teamed up with the OncoRNA lab of Prof. Mestdagh (University of Ghent), who has developed a cost-effective platform for parallelized shotgun transcriptomics, which offers high genome coverage. Together, we intend to deploy the combination of CP and transcriptomics for systematic gene silencing screens based on CRISPRi technology. As proof-of-concept, we will perform a targeted knockdown screen for a set of genes with known MoA in a panel of disease-relevant cell lines. By associating specific genes with simultaneous changes in cell morphology and gene expression profile, we aim to establish an enrichment analysis that allows unbiased MoA prediction. We will offer this platform as a service to biotechnology and pharmaceutical companies seeking to enhance their preclinical R&D lines. At the same time, we will build biological data capital, with which we intend to redesign the target discovery process and position ourselves in the vanguard of data-driven biotech at the European level.Researcher(s)
- Promoter: De Vos Winnok
- Co-promoter: Van De Looverbosch Tim
Research team(s)
Project type(s)
- Research Project