Loss of the heterogeneous expression of flippase ATP11B leads to cerebral small vessel disease in a normotensive rat model.

Cerebral small vessel disease (SVD) is the leading cause of vascular dementia, causes a quarter of strokes, and worsens stroke outcomes. The disease is characterised by patchy cerebral small vessel and white matter pathology, but the underlying mechanisms are poorly understood. This microvascular and tissue damage has been classically considered secondary to extrinsic factors, such as hypertension, but this fails to explain the patchy nature of the disease, the link to endothelial cell (EC) dysfunction even when hypertension is absent, and the increasing evidence of high heritability to SVD-related brain damage. We have previously shown the link between deletion of the phospholipase flippase Atp11b and EC dysfunction in an inbred hypertensive rat model with SVD-like pathology and a single nucleotide polymorphism (SNP) in ATP11B associated with human sporadic SVD. Here, we generated a novel normotensive transgenic rat model, where Atp11b is deleted, and show pathological, imaging and behavioural changes typical of those in human SVD, but that occur without hypertension. Atp11bKO rat brain and retinal small vessels show ECs with molecular and morphological changes of dysfunction, with myelin disruption in a patchy pattern around some but not all brain small vessels, similar to the human brain. We show that ATP11B/ATP11B is heterogeneously expressed in ECs in normal rat and human brain even in the same transverse section of the same blood vessel, suggesting variable effects of the loss of ATP11B on each vessel and an explanation for the patchy nature of the disease. This work highlights a link between inherent EC dysfunction and vulnerability to SVD white matter damage with a marked heterogeneity of ECs in vivo which modulates this response, occurring even in the absence of hypertension. These findings refocus our strategies for therapeutics away from antihypertensive (and vascular risk factor) control alone and towards ECs in the effort to provide alternative targets to prevent a major cause of stroke and dementia.

Detecting repeated cancer evolution from multi-region tumor sequencing data.

Recurrent successions of genomic changes, both within and between patients, reflect repeated evolutionary processes that are valuable for the anticipation of cancer progression. Multi-region sequencing allows the temporal order of some genomic changes in a tumor to be inferred, but the robust identification of repeated evolution across patients remains a challenge. We developed a machine-learning method based on transfer learning that allowed us to overcome the stochastic effects of cancer evolution and noise in data and identified hidden evolutionary patterns in cancer cohorts. When applied to multi-region sequencing datasets from lung, breast, renal, and colorectal cancer (768 samples from 178 patients), our method detected repeated evolutionary trajectories in subgroups of patients, which were reproduced in single-sample cohorts (n = 2,935). Our method provides a means of classifying patients on the basis of how their tumor evolved, with implications for the anticipation of disease progression.

Phylogenetic convolutional neural networks in metagenomics.

BACKGROUND: Convolutional Neural Networks can be effectively used only when data are endowed with an intrinsic concept of neighbourhood in the input space, as is the case of pixels in images. We introduce here Ph-CNN, a novel deep learning architecture for the classification of metagenomics data based on the Convolutional Neural Networks, with the patristic distance defined on the phylogenetic tree being used as the proximity measure. The patristic distance between variables is used together with a sparsified version of MultiDimensional Scaling to embed the phylogenetic tree in a Euclidean space. RESULTS: Ph-CNN is tested with a domain adaptation approach on synthetic data and on a metagenomics collection of gut microbiota of 38 healthy subjects and 222 Inflammatory Bowel Disease patients, divided in 6 subclasses. Classification performance is promising when compared to classical algorithms like Support Vector Machines and Random Forest and a baseline fully connected neural network, e.g. the Multi-Layer Perceptron. CONCLUSION: Ph-CNN represents a novel deep learning approach for the classification of metagenomics data. Operatively, the algorithm has been implemented as a custom Keras layer taking care of passing to the following convolutional layer not only the data but also the ranked list of neighbourhood of each sample, thus mimicking the case of image data, transparently to the user.

Competition for endothelial cell polarity drives vascular morphogenesis in the mouse retina.

Blood-vessel formation generates unique vascular patterns in each individual. The principles governing the apparent stochasticity of this process remain to be elucidated. Using mathematical methods, we find that the transition between two fundamental vascular morphogenetic programs-sprouting angiogenesis and vascular remodeling-is established by a shift of collective front-to-rear polarity of endothelial cells in the mouse retina. We demonstrate that the competition between biochemical (VEGFA) and mechanical (blood-flow-induced shear stress) cues controls this collective polarity shift. Shear stress increases tension at focal adhesions overriding VEGFA-driven collective polarization, which relies on tension at adherens junctions. We propose that vascular morphogenetic cues compete to regulate individual cell polarity and migration through tension shifts that translates into tissue-level emergent behaviors, ultimately leading to uniquely organized vascular patterns.

2D alpha-shapes to quantify retinal microvasculature morphology and their application to proliferative diabetic retinopathy characterisation in fundus photographs.

The use of 2D alpha-shapes (α-shapes) to quantify morphological features of the retinal microvasculature could lead to imaging biomarkers for proliferative diabetic retinopathy (PDR). We tested our approach using the MESSIDOR dataset that consists of colour fundus photographs from 547 healthy individuals, 149 with mild diabetic retinopathy (DR), 239 with moderate DR, 199 pre-PDR and 53 PDR. The skeleton (centrelines) of the automatically segmented retinal vasculature was represented as an α-shape and the proposed parameters, complexity ([Formula: see text]), spread (OpA), global shape (VS) and presence of abnormal angiogenesis (Gradα) were computed. In cross-sectional analysis, individuals with PDR had a lower [Formula: see text], OpA and Gradα indicating a vasculature that is more complex, less spread (i.e. dense) and the presence of numerous small vessels. The results show that α-shape parameters characterise vascular abnormalities predictive of PDR (AUC 0.73; 95% CI [0.73 0.74]) and have therefore potential to reveal changes in retinal microvascular morphology.

Automated Segmentation of Optical Coherence Tomography Angiography Images: Benchmark Data and Clinically Relevant Metrics.

Purpose: To generate the first open dataset of retinal parafoveal optical coherence tomography angiography (OCTA) images with associated ground truth manual segmentations, and to establish a standard for OCTA image segmentation by surveying a broad range of state-of-the-art vessel enhancement and binarization procedures. Methods: Handcrafted filters and neural network architectures were used to perform vessel enhancement. Thresholding methods and machine learning approaches were applied to obtain the final binarization. Evaluation was performed by using pixelwise metrics and newly proposed topological metrics. Finally, we compare the error in the computation of clinically relevant vascular network metrics (e.g., foveal avascular zone area and vessel density) across segmentation methods. Results: Our results show that, for the set of images considered, deep learning architectures (U-Net and CS-Net) achieve the best performance (Dice = 0.89). For applications where manually segmented data are not available to retrain these approaches, our findings suggest that optimally oriented flux (OOF) is the best handcrafted filter (Dice = 0.86). Moreover, our results show up to 25% differences in vessel density accuracy depending on the segmentation method used. Conclusions: In this study, we derive and validate the first open dataset of retinal parafoveal OCTA images with associated ground truth manual segmentations. Our findings should be taken into account when comparing the results of clinical studies and performing meta-analyses. Finally, we release our data and source code to support standardization efforts in OCTA image segmentation. Translational Relevance: This work establishes a standard for OCTA retinal image segmentation and introduces the importance of evaluating segmentation performance in terms of clinically relevant metrics.