Associations between yogurt intake and risk of diet-related cardiometabolic diseases (CMDs) have been the subject of recent research in epidemiologic nutrition. A healthy dietary pattern has been identified as a pillar for the prevention of weight gain and CMDs. Epidemiologic studies suggest that yogurt consumption is linked to healthy dietary patterns, lifestyles, and reduced risk of CMDs, particularly type 2 diabetes. However, to our knowledge, few to no randomized controlled trials have investigated yogurt intake in relation to cardiometabolic clinical outcomes. Furthermore, there has been little attempt to clarify the mechanisms that underlie the potential beneficial effects of yogurt consumption on CMDs. Yogurt is a nutrient-dense dairy food and has been suggested to reduce weight gain and prevent CMDs by contributing to intakes of protein, calcium, bioactive lipids, and several other micronutrients. In addition, fermentation with bacterial strains generates bioactive peptides, resulting in a potentially greater beneficial effect of yogurt on metabolic health than nonfermented dairy products such as milk. To date, there is little concrete evidence that the mechanisms proposed in observational studies to explain positive results of yogurt on CMDs or parameters are valid. Many proposed mechanisms are based on assumptions that commercial yogurts contain strain-specific probiotics, that viable yogurt cultures are present in adequate quantities, and that yogurt provides a minimum threshold dose of nutrients or bioactive components capable of exerting a physiologic effect. Therefore, the primary objective of this review is to investigate the plausibility of potential mechanisms commonly cited in the literature in order to shed light on the inverse associations reported between yogurt intake and various cardiometabolic health parameters that are related to its nutrient profile, bacterial constituents, and food matrix. This article reviews current gaps and challenges in identifying such mechanisms and provides a perspective on the research agenda to validate the proposed role of yogurt in protecting against CMDs.
Ala Little Melissa Model 36
While single-cell transcriptional profiling has greatly increased our capacity to interrogate biology, accurate cell classification within and between datasets is a key challenge. This is particularly so in pluripotent stem cell-derived organoids which represent a model of a developmental system. Here, clustering algorithms and selected marker genes can fail to accurately classify cellular identity while variation in analyses makes it difficult to meaningfully compare datasets. Kidney organoids provide a valuable resource to understand kidney development and disease. However, direct comparison of relative cellular composition between protocols has proved challenging. Hence, an unbiased approach for classifying cell identity is required.
The R package, scPred, was trained on multiple single cell RNA-seq datasets of human fetal kidney. A hierarchical model classified cellular subtypes into nephron, stroma and ureteric epithelial elements. This model, provided in the R package DevKidCC (github.com/KidneyRegeneration/DevKidCC), was then used to predict relative cell identity within published kidney organoid datasets generated using distinct cell lines and differentiation protocols, interrogating the impact of such variations. The package contains custom functions for the display of differential gene expression within cellular subtypes.
DevKidCC was used to directly compare between distinct kidney organoid protocols, identifying differences in relative proportions of cell types at all hierarchical levels of the model and highlighting variations in stromal and unassigned cell types, nephron progenitor prevalence and relative maturation of individual epithelial segments. Of note, DevKidCC was able to distinguish distal nephron from ureteric epithelium, cell types with overlapping profiles that have previously confounded analyses. When applied to a variation in protocol via the addition of retinoic acid, DevKidCC identified a consequential depletion of nephron progenitors.
Here we have taken reference HFK datasets from three publications that span multiple ages and kidney regions (Table 1), performed individual annotations of the cells present based on prior information, then used all confidently classified cells to train classification models using the R package scPred [53], a generalisable method which has showed high accuracy in different experiments and datasets from multiple tissues, and considered a top performer in benchmarking studies [9]. We finally utilise established knowledge of kidney developmental biology to refine the classification of off-target cell types. The resulting model, referred to as DevKidCC, provides a robust and accurate classification of cells in novel single-cell datasets generated from developing human kidney or stem cell-derived kidney organoids. DevKidCC defines a model of cellular identity organised in a hierarchical manner to represent the key developmental trajectories of lineages within the developing kidney. The classification method is complemented with custom visualization tools in the DevKidCC package. This classifier was then used to investigate published kidney organoid datasets to compare organoid patterning and gene expression profiles across these datasets. We present a variety of applications of DevKidCC to the reanalysis of existing data. This analysis revealed differences in cell type proportions, nephron patterning and maturation between kidney organoid protocols. We also applied DevKidCC to investigate approaches for directed differentiation to one cell population, the ureteric epithelium, and dissect the effect of all-trans retinoic acid on nephron patterning and podocyte maturation. While DevKidCC is specifically trained on HFK for application to kidney organoid models, the development framework presented here could be applied for any tissue system to generate a cell classification model.
We generated a comprehensive developing kidney reference single-cell dataset by harmonising the raw data from multiple high-quality human fetal kidney datasets. The annotation of the reference included three tiers with increasing specificity, with a clear hierarchical structure between the tiers. This dataset was then used to train machine learning models using the R package scPred [53]. One model was trained for each node of identities within the classification hierarchy.
Utilising scPred [53] the classifiers were trained using the same parameters, with the relevant cells inputted for each. The feature space used was the top 100 principal components. The classifiers were trained using a support vector machine with a radial kernel using one round of harmonisation. The classifiers are stored as a scPred [53]object and can be used to classify cells within a Seurat [7, 8] object using the scPred [53] package. These classifiers will calculate the probability of a cell belonging to the trained identities within that classifier, giving a probability score between 0 and 1 for each identity. It will then assign an identity of the highest score above the set threshold or call the cell unassigned if no identity scores above the threshold. Classification is organised in a biologically relevant hierarchy so as to optimally and accurately identify the cellular identity of all analysed cells. All cells are first classified using the first-tier model, containing generalised lineage identities of stroma, nephron progenitors, nephron, ureteric epithelium and endothelium. After probability calculation using the first-tier model, cells that do not pass the threshold are classified as unassigned. The area under the AUROC and AUPRC to decide a threshold were determined using the MLeval R package from CRAN -project.org/web/packages/MLeval/index.html [54]. The threshold is set to 0.7 by default but can be adjusted by the user, which can be useful if the user wants to classify cells with decreasing degrees of probability. NPC cells were subjected to further investigation by subsetting and reclustering at a resolution level of 0.5 using FindClusters and identifying the percentage of cells expressing PAX2 with clusters below 30% being relabelled as NPC-like. Cells assigned to stroma, nephron and ureteric epithelium are passed into a second tier of classification specific to these identities. It is important to note that at the second and third classification tiers, there is no thresholding, i.e., all cells are assigned an identity with no cells classed as unassigned. The second-tier ureteric epithelium model is trained on the tip, cortical, outer and inner medullary cell identities. The second-tier stroma model is trained on the stromal progenitors, cortex, medullary and mesangial cell identities. The second-tier nephron model is trained on the early nephron, distal nephron, proximal nephron, renal corpuscle and nephron cell cycle population. The distal nephron, proximal nephron and renal corpuscle are then further classified into more specific identities in a third tier of models. The third-tier distal nephron model is trained on early distal/medial cells, distal tubule and loop of Henle cells. The third-tier proximal nephron model is trained on early proximal tubule and proximal tubule cells. The third-tier renal corpuscle model is trained on parietal epithelial cells, early podocytes and podocytes. Each stage of the classification step is recorded as a metadata column, as is the final classification for each cell. All the probability scores and tier classifications are readily accessible within the Seurat [7, 8] object for further analysis.
Generation of a comprehensive reference to train classification models. A UMAP visualisation of the integrated reference HFK datasets. B Expression of marker genes in the integrated reference shown by annotated identity. C Graphical representation of the DevKidCC model hierarchy and classification process. HFK human fetal kidney, Pct. percent of, Exp. expression 2ff7e9595c
Comments