Comparison of Artificial Intelligence based approaches to cell function prediction

Integrating Artificial Intelligence in the Diagnosis and Management of Metabolic Syndrome: A Comprehensive Review

BACKGROUND

Metabolic syndrome (MetS) is a progressive chronic pathophysiological state characterised by abdominal obesity, hypertension, hyperglycaemia, and dyslipidaemia. It is recognised as one of the major clinical syndromes affecting human health, with approximately one-quarter of the global population impacted. MetS increases the risk of developing cardiovascular diseases (CVDs), stroke, type 2 diabetes mellitus (T2DM), and diverse metabolic diseases. Early diagnosis of MetS could potentially reduce the prevalence of these diseases. However, care for the MetS population faces significant challenges due to (i) a lack of comprehensive understanding of the full spectrum of associated diseases, stemming from unclear pathophysiological mechanisms and (ii) frequent underdiagnosis or misdiagnosis of MetS in clinical settings due to inconsistent screening guidelines, limited medical resources, time constraints in clinical practice, and insufficient awareness and training. The increasing availability of healthcare and medical data presents opportunities to apply and innovate with artificial intelligence (AI) in addressing these challenges. This review aims to (i) summarise the spectrum of diseases associated with MetS and (ii) review the diverse AI models applied to MetS and metabolic syndrome-related diseases (MetSRD), where MetSRD collectively refers to diseases and conditions directly associated with MetS.

METHODS

Our review consists of two phases. Initially, we conducted a literature review on MetS to narrow down the spectrum of MetSRD based on the strength of clinical evidence. We then used the terms 'Metabolic Syndrome' and 'Machine Learning' in combination with the identified MetSRD for further refinement. In total, we identified 52 related diseases in the first phase and 36 articles in the second phase.

RESULTS

We identified a total of 52 MetSRD after the first phase, with T2DM, CVDs, and cancer being the top three. Among the 36 articles obtained in the second phase, we observed the following: (i) The criteria for MetS were inconsistent across the studies. (ii) The primary purpose of AI applications was to identify risk factors for diseases, thereby improving predictions for MetS or MetSRD. Traditional machine learning models, such as Random Forest and Logistic Regression, were found to be the most effective. (iii) In addition to the MetS criteria, AI models explored other factors, including demographic and physiological variables, dietary influences, lipidomic and proteomic indicators, and more.

CONCLUSION

This review underscores the significant link between MetS and a spectrum of diseases, with a particular focus on underreported conditions such as non-alcoholic fatty liver disease and stroke. Through the analysis of data from diverse sources, AI models, and MetS diagnostic criteria, additional indicators beyond traditional measures have been identified, emphasising the importance of combining both traditional and non-traditional markers to enhance the diagnostic and predictive capabilities for MetS and MetSRD. AI shows great potential in MetS research, particularly through the integration of multi-source data, including clinical metrics, genetic information, and omics data. The amalgamation of traditional machine learning and modern machine learning models is particularly promising, offering a balanced approach to model performance and data complexity. While international definitions provide global applicability, they may not be suitable for all populations and scenarios, necessitating flexible diagnostic criteria and adaptive, explainable algorithms. Ultimately, these will enable personalised diagnostics and targeted interventions.

Recent Progress in Retinal Pigment Epithelium Cell-Based Therapy for Retinal Disease

Age-related macular degeneration and retinitis pigmentosa are degenerative retinal diseases that cause severe vision loss. Early clinical trials involving transplantation of retinal pigment epithelial cells and/or photoreceptors as a treatment for these conditions are underway. In this review, we summarize recent progress in the field of retinal pigment epithelium transplantation, including some pertinent clinical trial results as well as preclinical studies that address issues of transplant immunology, cell delivery, and cell manufacturing.

Operationalizing the Use of Biofabricated Tissue Models as Preclinical Screening Platforms for Drug Discovery and Development

A wide range of complex in vitro models (CIVMs) are being developed for scientific research and preclinical drug efficacy and safety testing. The hope is that these CIVMs will mimic human physiology and pathology and predict clinical responses more accurately than the current cellular models. The integration of these CIVMs into the drug discovery and development pipeline requires rigorous scientific validation, including cellular, morphological, and functional characterization; benchmarking of clinical biomarkers; and operationalization as robust and reproducible screening platforms. It will be critical to establish the degree of physiological complexity that is needed in each CIVM to accurately reproduce native-like homeostasis and disease phenotypes, as well as clinical pharmacological responses. Choosing which CIVM to use at each stage of the drug discovery and development pipeline will be driven by a fit-for-purpose approach, based on the specific disease pathomechanism to model and screening throughput needed. Among the different CIVMs, biofabricated tissue equivalents are emerging as robust and versatile cellular assay platforms. Biofabrication technologies, including bioprinting approaches with hydrogels and biomaterials, have enabled the production of tissues with a range of physiological complexity and controlled spatial arrangements in multiwell plate platforms, which make them amenable for medium-throughput screening. However, operationalization of such 3D biofabricated models using existing automation screening platforms comes with a unique set of challenges. These challenges will be discussed in this perspective, including examples and thoughts coming from a laboratory dedicated to designing and developing assays for automated screening.

Reproducible production and image-based quality evaluation of retinal pigment epithelium sheets from human induced pluripotent stem cells

Transplantation of retinal pigment epithelial (RPE) sheets derived from human induced pluripotent cells (hiPSC) is a promising cell therapy for RPE degeneration, such as in age-related macular degeneration. Current RPE replacement therapies, however, face major challenges. They require a tedious manual process of selecting differentiated RPE from hiPSC-derived cells, and despite wide variation in quality of RPE sheets, there exists no efficient process for distinguishing functional RPE sheets from those unsuitable for transplantation. To overcome these issues, we developed methods for the generation of RPE sheets from hiPSC, and image-based evaluation. We found that stepwise treatment with six signaling pathway inhibitors along with nicotinamide increased RPE differentiation efficiency (RPE6iN), enabling the RPE sheet generation at high purity without manual selection. Machine learning models were developed based on cellular morphological features of F-actin-labeled RPE images for predicting transepithelial electrical resistance values, an indicator of RPE sheet function. Our model was effective at identifying low-quality RPE sheets for elimination, even when using label-free images. The RPE6iN-based RPE sheet generation combined with the non-destructive image-based prediction offers a comprehensive new solution for the large-scale production of pure RPE sheets with lot-to-lot variations and should facilitate the further development of RPE replacement therapies.