Deep Learning-based Image Analysis for High-content Screening

<p><b>High-content screening is an empirical strategy in drug discovery toidentify substances capable of altering cellular phenotype — the set ofobservable characteristics of a cell — in a desired way. Throughout thepast two decades, high-content screening has gathered significant attentionfrom academia and the pharmaceutical industry. However, imageanalysis remains a considerable hindrance to the widespread applicationof high-content screening. Standard image analysis relies on feature engineeringand suffers from inherent drawbacks such as the dependence onannotated inputs. There is an urging need for reliable and more efficientmethods to cope with increasingly large amounts of data produced.</b></p> <p>This thesis centres around the design and implementation of a deeplearning-based image analysis pipeline for high-content screening. Theend goal is to identify and cluster hit compounds that significantly alterthe phenotype of a cell. The proposed pipeline replaces feature engineeringwith a k-nearest neighbour-based similarity analysis. In addition, featureextraction using convolutional autoencoders is applied to reduce thenegative effects of noise on hit selection. As a result, the feature engineeringprocess is circumvented. A novel similarity measure is developed tofacilitate similarity analysis. Moreover, we combine deep learning withstatistical modelling to achieve optimal results. Preliminary explorationssuggest that the choice of hyperparameters have a direct impact on neuralnetwork performance. Generalised estimating equation models are usedto predict the most suitable neural network architecture for the input data.</p> <p>Using the proposed pipeline, we analyse an extensive set of images acquiredfrom a series of cell-based assays examining the effect of 282 FDAapproved drugs. The analysis of this data set produces a shortlist of drugsthat can significantly alter a cell’s phenotype, then further identifies fiveclusters of the shortlisted drugs. The clustering results present groups ofexisting drugs that have the potential to be repurposed for new therapeuticuses. Furthermore, our findings align with published studies. Comparedwith other neural networks, the image analysis pipeline proposedin this thesis provides reliable and better results in a shorter time frame.</p>

Deep Learning-based Image Analysis for High-content Screening

<p><b>High-content screening is an empirical strategy in drug discovery toidentify substances capable of altering cellular phenotype — the set ofobservable characteristics of a cell — in a desired way. Throughout thepast two decades, high-content screening has gathered significant attentionfrom academia and the pharmaceutical industry. However, imageanalysis remains a considerable hindrance to the widespread applicationof high-content screening. Standard image analysis relies on feature engineeringand suffers from inherent drawbacks such as the dependence onannotated inputs. There is an urging need for reliable and more efficientmethods to cope with increasingly large amounts of data produced.</b></p> <p>This thesis centres around the design and implementation of a deeplearning-based image analysis pipeline for high-content screening. Theend goal is to identify and cluster hit compounds that significantly alterthe phenotype of a cell. The proposed pipeline replaces feature engineeringwith a k-nearest neighbour-based similarity analysis. In addition, featureextraction using convolutional autoencoders is applied to reduce thenegative effects of noise on hit selection. As a result, the feature engineeringprocess is circumvented. A novel similarity measure is developed tofacilitate similarity analysis. Moreover, we combine deep learning withstatistical modelling to achieve optimal results. Preliminary explorationssuggest that the choice of hyperparameters have a direct impact on neuralnetwork performance. Generalised estimating equation models are usedto predict the most suitable neural network architecture for the input data.</p> <p>Using the proposed pipeline, we analyse an extensive set of images acquiredfrom a series of cell-based assays examining the effect of 282 FDAapproved drugs. The analysis of this data set produces a shortlist of drugsthat can significantly alter a cell’s phenotype, then further identifies fiveclusters of the shortlisted drugs. The clustering results present groups ofexisting drugs that have the potential to be repurposed for new therapeuticuses. Furthermore, our findings align with published studies. Comparedwith other neural networks, the image analysis pipeline proposedin this thesis provides reliable and better results in a shorter time frame.</p>

Organoids in image-based phenotypic chemical screens

Image-based phenotypic screening relies on the extraction of multivariate information from cells cultured under a large variety of conditions. Technical advances in high-throughput microscopy enable screening in increasingly complex and biologically relevant model systems. To this end, organoids hold great potential for high-content screening because they recapitulate many aspects of parent tissues and can be derived from patient material. However, screening is substantially more difficult in organoids than in classical cell lines from both technical and analytical standpoints. In this review, we present an overview of studies employing organoids for screening applications. We discuss the promises and challenges of small-molecule treatments in organoids and give practical advice on designing, running, and analyzing high-content organoid-based phenotypic screens.

A Novel High-Content Screening-Based Method for Anti-Trypanosoma cruzi Drug Discovery Using Human-Induced Pluripotent Stem Cell-Derived Cardiomyocytes

Chagas disease is caused by Trypanosoma cruzi infection and remains a relevant cause of chronic heart failure in Latin America. The pharmacological arsenal for Chagas disease is limited, and the available anti-T. cruzi drugs are not effective when administered during the chronic phase. Cardiomyocytes derived from human-induced pluripotent stem cells (hiPSC-CMs) have the potential to accelerate the process of drug discovery for Chagas disease, through predictive preclinical assays in target human cells. Here, we aimed to establish a novel high-content screening- (HCS-) based method using hiPSC-CMs to simultaneously evaluate anti-T. cruzi activity and cardiotoxicity of chemical compounds. To provide proof-of-concept data, the reference drug benznidazole and three compounds with known anti-T. cruzi activity (a betulinic acid derivative named BA5 and two thiazolidinone compounds named GT5A and GT5B) were evaluated in the assay. hiPSC-CMs were infected with T. cruzi and incubated for 48 h with serial dilutions of the compounds for determination of EC50 and CC50 values. Automated multiparametric analyses were performed using an automated high-content imaging system. Sublethal toxicity measurements were evaluated through morphological measurements related to the integrity of the cytoskeleton by phalloidin staining, nuclear score by Hoechst 33342 staining, mitochondria score following MitoTracker staining, and quantification of NT-pro-BNP, a peptide released upon mechanical myocardial stress. The compounds showed EC50 values for anti-T. cruzi activity similar to those previously described for other cell types, and GT5B showed a pronounced trypanocidal activity in hiPSC-CMs. Sublethal changes in cytoskeletal and nucleus scores correlated with NT-pro-BNP levels in the culture supernatant. Mitochondrial score changes were associated with increased cytotoxicity. The assay was feasible and allowed rapid assessment of anti-T. cruzi action of the compounds, in addition to cardiotoxicity parameters. The utilization of hiPSC-CMs in the drug development workflow for Chagas disease may help in the identification of novel compounds.

Apoptosis assessment in high-content and high-throughput screening assays

Here the authors describe the development of AUTOptosis, an economical and rapid apoptosis monitoring method suitable for high-content and high-throughput screening assays. AUTOptosis is based on the quantification of nuclei intensity via staining with Hoechst 33342. First, the authors calibrated the method using standard apoptosis inducers in multiple cell lines. Next, the authors validated the applicability of this approach to high-content screening using a small library of compounds and compared it with the terminal deoxynucleotidyl transferase dUTP nick end labeling gold standard. Finally, the authors demonstrated the specificity of the method by using AUTOposis to detect apoptosis triggered by Mycobacterium tuberculosis intracellular infections.