Search

Current liver fibrosis scoring by computer-assisted image analytics is not fully automated as it requires manual preprocessing (segmentation and feature extraction) typically based on domain knowledge in liver pathology. Deep learning-based algorithms can potentially classify these images without the need for preprocessing through learning from a large dataset of images. We investigated the performance of classification models built using a deep learning-based algorithm pre-trained using multiple sources of images to score liver fibrosis and compared them against conventional non-deep learning-based algorithms - artificial neural networks (ANN), multinomial logistic regression (MLR), support vector machines (SVM) and random forests (RF). Automated feature classification and fibrosis scoring were achieved by using a transfer learning-based deep learning network, AlexNet-Convolutional Neural Networks (CNN), with balanced area under receiver operating characteristic (AUROC) values of up to 0.85–0.95 versus ANN (AUROC of up to 0.87–1.00), MLR (AUROC of up to 0.73–1.00), SVM (AUROC of up to 0.69–0.99) and RF (AUROC of up to 0.94–0.99). Results indicate that a deep learning-based algorithm with transfer learning enables the construction of a fully automated and accurate prediction model for scoring liver fibrosis stages that is comparable to other conventional non-deep learning-based algorithms that are not fully automated.

This chapter will cover the use of stem cells and tissue engineering technologies in the assessment of developmental and reproductive toxicity (DART). We will first give an overview of current animal-based DART regulatory guidelines to highlight opportunities where in vitro assays can replace or complement animal testing in the identification of potential teratogens. To date, animal embryo-based in vitro assays, such as rat whole embryo culture (WEC), zebrafish and chicken embryo models, have been explored as animal alternatives for teratogen testing. We will briefly summarize salient features of these in vitro embryo culture systems and discuss recent technological advances, such as the use of microfluidic platforms to facilitate improvements in assay robustness and throughput. More recently, pluripotent stem-cell (PSC)-based assays for teratogen screening are increasingly being developed because they can potentially be human-specific via the use of human PSCs. Moreover, PSC-based models allow one to probe and measure teratogenic effects through multiple readouts, including changes in molecular biomarker expression, metabolite profiles, as well as cellular differentiation and migration phenotypes. The adoption of tissue engineering approaches to control PSC differentiation rates into various tissue types will allow for more realistic modeling of embryonic developmental processes to screen for potential teratogens with higher throughput and precision.

Idiosyncratic drug-induced hepatotoxicity is a major cause of liver damage and drug pipeline failure, and is difficult to study as patient-specific features are not readily incorporated in traditional hepatotoxicity testing approaches using population pooled cell sources. Here we demonstrate the use of patient-specific hepatocyte-like cells (HLCs) derived from induced pluripotent stem cells for modeling idiosyncratic hepatotoxicity to pazopanib (PZ), a tyrosine kinase inhibitor drug associated with significant hepatotoxicity of unknown mechanistic basis. In vitro cytotoxicity assays confirmed that HLCs from patients with clinically identified hepatotoxicity were more sensitive to PZ-induced toxicity than other individuals, while a prototype hepatotoxin acetaminophen was similarly toxic to all HLCs studied. Transcriptional analyses showed that PZ induces oxidative stress (OS) in HLCs in general, but in HLCs from susceptible individuals, PZ causes relative disruption of iron metabolism and higher burden of OS. Our study establishes the first patient-specific HLC-based platform for idiosyncratic hepatotoxicity testing, incorporating multiple potential causative factors and permitting the correlation of transcriptomic and cellular responses to clinical phenotypes. Establishment of patient-specific HLCs with clinical phenotypes representing population variations will be valuable for pharmaceutical drug testing. © The Author(s) 2017.

Heterogeneity in human pluripotent stem cell (PSC) fates is partially caused by mechanical asymmetry arising from spatial polarization of cell-cell and cell-matrix adhesions. Independent studies have shown that integrin and E-cadherin adhesions promote opposing differentiation and pluripotent fates respectively although their crosstalk mechanism in modulating cell fate heterogeneity remains unknown. Here, we demonstrated that spatial polarization of integrin and E-cadherin adhesions in a human PSC colony compete to recruit Rho-ROCK activated myosin II to different localities to pattern pluripotent-differentiation decisions, resulting in spatially heterogeneous colonies. Cell micropatterning was used to modulate the spatial polarization of cell adhesions, which enabled us to prospectively determine localization patterns of activated myosin II and mesoendoderm differentiation. Direct inhibition of Rho-ROCK-myosin II activation phenocopied E-cadherin rather than integrin inhibition to form uniformly differentiated colonies. This indicated that E-cadherin was the primary gatekeeper to differentiation progression. This insight allows for biomaterials to be tailored for human PSC maintenance or differentiation with minimal heterogeneity.

Unintended exposure to teratogenic compounds can lead to various birth defects; however current animal-based testing is limited by time, cost and high inter-species variability. Here, we developed a human-relevant in vitro model, which recapitulated two cellular events characteristic of embryogenesis, to identify potentially teratogenic compounds. We spatially directed mesoendoderm differentiation, epithelial-mesenchymal transition and the ensuing cell migration in micropatterned human pluripotent stem cell (hPSC) colonies to collectively form an annular mesoendoderm pattern. Teratogens could disrupt the two cellular processes to alter the morphology of the mesoendoderm pattern. Image processing and statistical algorithms were developed to quantify and classify the compounds' teratogenic potential. We not only could measure dose-dependent effects but also correctly classify species-specific drug (Thalidomide) and false negative drug (D-penicillamine) in the conventional mouse embryonic stem cell test. This model offers a scalable screening platform to mitigate the risks of teratogen exposures in human.

Unintended exposure to teratogenic compounds can lead to various birth defects; however current animal-based testing is limited by time, cost and high inter-species variability. Here, we developed a human-relevant in vitro model, which recapitulated two cellular events characteristic of embryogenesis, to identify potentially teratogenic compounds. We spatially directed mesoendoderm differentiation, epithelial-mesenchymal transition and the ensuing cell migration in micropatterned human pluripotent stem cell (hPSC) colonies to collectively form an annular mesoendoderm pattern. Teratogens could disrupt the two cellular processes to alter the morphology of the mesoendoderm pattern. Image processing and statistical algorithms were developed to quantify and classify the compounds’ teratogenic potential. We not only could measure dose-dependent effects but also correctly classify species-specific drug (Thalidomide) and false negative drug (D-penicillamine) in the conventional mouse embryonic stem cell test. This model offers a scalable screening platform to mitigate the risks of teratogen exposures in human., Singapore. Agency for Science, Technology and Research, Janssen Pharmaceutical Ltd. (Grant R-185-000-182-592), Janssen Pharmaceutical Ltd. (Grant R-185-000-228-592), Singapore-MIT Alliance Computational and Systems Biology Flagship Project (C-382-641-001-091), Mechanobiology Institute, Singapore (R-714-001-003-271)

Human embryonic stem cell (hESC)-based developmental toxicity models can help to reduce animal testing and mitigate inter-species variability. However, current hESC models do not fully recapitulate embryonic developmental structural changes, which are underpinned by spatio-temporally controlled differentiation and migration events. We show that mechanical polarization imposed by cell micropatterning could spatially pattern hESC differentiation and migration, and in turn reconstituted a rudimentary structural motif. Features of this structural motif were disrupted by paradigm developmental toxins but not a negative control drug. This novel micropatterned hESC model can potentially be scaled- up into an array format for high throughput developmental toxicity screening.