SpatialDDLS: an R package to deconvolute spatial transcriptomics data using neural networks

SUMMARY

Spatial transcriptomics has changed our way to study tissue structure and cellular organization. However, there are still limitations in its resolution, and most available platforms do not reach a single cell resolution. To address this issue, we introduce SpatialDDLS, a fast neural network-based algorithm for cell type deconvolution of spatial transcriptomics data. SpatialDDLS leverages single-cell RNA sequencing data to simulate mixed transcriptional profiles with predefined cellular composition, which are subsequently used to train a fully connected neural network to uncover cell type diversity within each spot. By comparing it with two state-of-the-art spatial deconvolution methods, we demonstrate that SpatialDDLS is an accurate and fast alternative to the available state-of-the art tools.

AVAILABILITY AND IMPLEMENTATION

The R package SpatialDDLS is available via CRAN-The Comprehensive R Archive Network: https://CRAN.R-project.org/package=SpatialDDLS. A detailed manual of the main functionalities implemented in the package can be found at https://diegommcc.github.io/SpatialDDLS.

A deep learning-based tool for the automated detection and analysis of caveolae in transmission electron microscopy images

Caveolae are nanoscopic and mechanosensitive invaginations of the plasma membrane, essential for adipocyte biology. Transmission electron microscopy (TEM) offers the highest resolution for caveolae visualization, but provides complicated images that are difficult to classify or segment using traditional automated algorithms such as threshold-based methods. As a result, the time-consuming tasks of localization and quantification of caveolae are currently performed manually. We used the Keras library in R to train a convolutional neural network with a total of 36,000 TEM image crops obtained from adipocytes previously annotated manually by an expert. The resulting model can differentiate caveolae from non-caveolae regions with a 97.44% accuracy. The predictions of this model are further processed to obtain caveolae central coordinate detection and cytoplasm boundary delimitation. The model correctly finds negligible caveolae predictions in images from caveolae depleted Cav1-/- adipocytes. In large reconstructions of adipocyte sections, model and human performances are comparable. We thus provide a new tool for accurate caveolae automated analysis that could speed up and assist in the characterization of the cellular mechanical response.

Chromatin dynamics through mouse preimplantation development revealed by single molecule localisation microscopy

Most studies addressing chromatin behaviour during preimplantation development are based on biochemical assays that lack spatial and cell-specific information, crucial during early development. Here, we describe the changes in chromatin taking place at the transition from totipotency to lineage specification, by using direct stochastical optical reconstruction microscopy (dSTORM) in whole-mount embryos during the first stages of mouse development. Through the study of two post-translational modifications of Histone 3 related to active and repressed chromatin, H3K4me3 and H3K9me3 respectively, we obtained a time-course of chromatin states, showing spatial differences between cell types, related to their differentiation state. This analysis adds a new layer of information to previous biochemical studies and provides novel insight to current models of chromatin organisation during the first stages of development.

SUMMARY: We have applied super-resolution microscopy to analyse changes in the state of chromatin during the first stages of mouse development, from the two-cell stage to the blastocyst.

Von Hippel-Lindau Protein Is Required for Optimal Alveolar Macrophage Terminal Differentiation, Self-Renewal, and Function

The rapid transit from hypoxia to normoxia in the lung that follows the first breath in newborn mice coincides with alveolar macrophage (AM) differentiation. However, whether sensing of oxygen affects AM maturation and function has not been previously explored. We have generated mice whose AMs show a deficient ability to sense oxygen after birth by deleting Vhl, a negative regulator of HIF transcription factors, in the CD11c compartment (CD11cΔVhl mice). VHL-deficient AMs show an immature-like phenotype and an impaired self-renewal capacity in vivo that persists upon culture ex vivo. VHL-deficient phenotype is intrinsic in AMs derived from monocyte precursors in mixed bone marrow chimeras. Moreover, unlike control Vhl^fl/fl, AMs from CD11cΔVhl mice do not reverse pulmonary alveolar proteinosis when transplanted into Csf2rb^−/− mice, demonstrating that VHL contributes to AM-mediated surfactant clearance. Thus, our results suggest that optimal AM terminal differentiation, self-renewal, and homeostatic function requires their intact oxygen-sensing capacity.

•

Hypoxia and glycolysis signatures are downregulated during alveolar macrophage maturation

•

VHL-deficient alveolar macrophages induce HIF-target genes and are more glycolytic

•

VHL-deficient alveolar macrophages show an intrinsic immature-like phenotype

•

VHL promotes alveolar macrophage self-renewal and surfactant clearance capacity

Instantaneous and non-destructive relative water content estimation from deep learning applied to resonant ultrasonic spectra of plant leaves

BACKGROUND

Non-contact resonant ultrasound spectroscopy (NC-RUS) has been proven as a reliable technique for the dynamic determination of leaf water status. It has been already tested in more than 50 plant species. In parallel, relative water content (RWC) is highly used in the ecophysiological field to describe the degree of water saturation in plant leaves. Obtaining RWC implies a cumbersome and destructive process that can introduce artefacts and cannot be determined instantaneously.

RESULTS

Here, we present a method for the estimation of RWC in plant leaves from non-contact resonant ultrasound spectroscopy (NC-RUS) data. This technique enables to collect transmission coefficient in a [0.15-1.6] MHz frequency range from plant leaves in a non-invasive, non-destructive and rapid way. Two different approaches for the proposed method are evaluated: convolutional neural networks (CNN) and random forest (RF). While CNN takes the entire ultrasonic spectra acquired from the leaves, RF only uses four relevant parameters resulted from the transmission coefficient data. Both methods were tested successfully in Viburnum tinus leaf samples with Pearson's correlations between 0.92 and 0.84.

CONCLUSIONS

This study showed that the combination of NC-RUS technique with deep learning algorithms is a robust tool for the instantaneous, accurate and non-destructive determination of RWC in plant leaves.

Flow Cytometry Data Preparation Guidelines for Improved Automated Phenotypic Analysis

Advances in flow cytometry (FCM) increasingly demand adoption of computational analysis tools to tackle the ever-growing data dimensionality. In this study, we tested different data input modes to evaluate how cytometry acquisition configuration and data compensation procedures affect the performance of unsupervised phenotyping tools. An analysis workflow was set up and tested for the detection of changes in reference bead subsets and in a rare subpopulation of murine lymph node CD103⁺ dendritic cells acquired by conventional or spectral cytometry. Raw spectral data or pseudospectral data acquired with the full set of available detectors by conventional cytometry consistently outperformed datasets acquired and compensated according to FCM standards. Our results thus challenge the paradigm of one-fluorochrome/one-parameter acquisition in FCM for unsupervised cluster-based analysis. Instead, we propose to configure instrument acquisition to use all available fluorescence detectors and to avoid integration and compensation procedures, thereby using raw spectral or pseudospectral data for improved automated phenotypic analysis.

Pulmonary Artery-Vein Classification in CT Images Using Deep Learning

Recent studies show that pulmonary vascular diseases may specifically affect arteries or veins through different physiologic mechanisms. To detect changes in the two vascular trees, physicians manually analyze the chest computed tomography (CT) image of the patients in search of abnormalities. This process is time consuming, difficult to standardize, and thus not feasible for large clinical studies or useful in real-world clinical decision making. Therefore, automatic separation of arteries and veins in CT images is becoming of great interest, as it may help physicians to accurately diagnose pathological conditions. In this paper, we present a novel, fully automatic approach to classify vessels from chest CT images into arteries and veins. The algorithm follows three main steps: first, a scale-space particles segmentation to isolate vessels; then a 3-D convolutional neural network (CNN) to obtain a first classification of vessels; finally, graph-cuts' optimization to refine the results. To justify the usage of the proposed CNN architecture, we compared different 2-D and 3-D CNNs that may use local information from bronchus- and vessel-enhanced images provided to the network with different strategies. We also compared the proposed CNN approach with a random forests (RFs) classifier. The methodology was trained and evaluated on the superior and inferior lobes of the right lung of 18 clinical cases with noncontrast chest CT scans, in comparison with manual classification. The proposed algorithm achieves an overall accuracy of 94%, which is higher than the accuracy obtained using other CNN architectures and RF. Our method was also validated with contrast-enhanced CT scans of patients with chronic thromboembolic pulmonary hypertension to demonstrate that our model generalizes well to contrast-enhanced modalities. The proposed method outperforms state-of-the-art methods, paving the way for future use of 3-D CNN for artery/vein classification in CT images.

Tox_(R)CNN: Deep learning-based nuclei profiling tool for drug toxicity screening

Toxicity is an important factor in failed drug development, and its efficient identification and prediction is a major challenge in drug discovery. We have explored the potential of microscopy images of fluorescently labeled nuclei for the prediction of toxicity based on nucleus pattern recognition. Deep learning algorithms obtain abstract representations of images through an automated process, allowing them to efficiently classify complex patterns, and have become the state-of-the art in machine learning for computer vision. Here, deep convolutional neural networks (CNN) were trained to predict toxicity from images of DAPI-stained cells pre-treated with a set of drugs with differing toxicity mechanisms. Different cropping strategies were used for training CNN models, the nuclei-cropping-based Tox_CNN model outperformed other models classifying cells according to health status. Tox_CNN allowed automated extraction of feature maps that clustered compounds according to mechanism of action. Moreover, fully automated region-based CNNs (RCNN) were implemented to detect and classify nuclei, providing per-cell toxicity prediction from raw screening images. We validated both Tox_(R)CNN models for detection of pre-lethal toxicity from nuclei images, which proved to be more sensitive and have broader specificity than established toxicity readouts. These models predicted toxicity of drugs with mechanisms of action other than those they had been trained for and were successfully transferred to other cell assays. The Tox_(R)CNN models thus provide robust, sensitive, and cost-effective tools for in vitro screening of drug-induced toxicity. These models can be adopted for compound prioritization in drug screening campaigns, and could thereby increase the efficiency of drug discovery.

Automatic Synthesis of Anthropomorphic Pulmonary CT Phantoms

The great density and structural complexity of pulmonary vessels and airways impose limitations on the generation of accurate reference standards, which are critical in training and in the validation of image processing methods for features such as pulmonary vessel segmentation or artery-vein (AV) separations. The design of synthetic computed tomography (CT) images of the lung could overcome these difficulties by providing a database of pseudorealistic cases in a constrained and controlled scenario where each part of the image is differentiated unequivocally. This work demonstrates a complete framework to generate computational anthropomorphic CT phantoms of the human lung automatically. Starting from biological and image-based knowledge about the topology and relationships between structures, the system is able to generate synthetic pulmonary arteries, veins, and airways using iterative growth methods that can be merged into a final simulated lung with realistic features. A dataset of 24 labeled anthropomorphic pulmonary CT phantoms were synthesized with the proposed system. Visual examination and quantitative measurements of intensity distributions, dispersion of structures and relationships between pulmonary air and blood flow systems show good correspondence between real and synthetic lungs (p > 0.05 with low Cohen's d effect size and AUC values), supporting the potentiality of the tool and the usefulness of the generated phantoms in the biomedical image processing field.

Comparing algorithms for automated vessel segmentation in computed tomography scans of the lung: the VESSEL12 study

The VESSEL12 (VESsel SEgmentation in the Lung) challenge objectively compares the performance of different algorithms to identify vessels in thoracic computed tomography (CT) scans. Vessel segmentation is fundamental in computer aided processing of data generated by 3D imaging modalities. As manual vessel segmentation is prohibitively time consuming, any real world application requires some form of automation. Several approaches exist for automated vessel segmentation, but judging their relative merits is difficult due to a lack of standardized evaluation. We present an annotated reference dataset containing 20 CT scans and propose nine categories to perform a comprehensive evaluation of vessel segmentation algorithms from both academia and industry. Twenty algorithms participated in the VESSEL12 challenge, held at International Symposium on Biomedical Imaging (ISBI) 2012. All results have been published at the VESSEL12 website http://vessel12.grand-challenge.org. The challenge remains ongoing and open to new participants. Our three contributions are: (1) an annotated reference dataset available online for evaluation of new algorithms; (2) a quantitative scoring system for objective comparison of algorithms; and (3) performance analysis of the strengths and weaknesses of the various vessel segmentation methods in the presence of various lung diseases.

A benchmark for comparison of cell tracking algorithms

Motivation: Automatic tracking of cells in multidimensional time-lapse fluorescence microscopy is an important task in many biomedical applications. A novel framework for objective evaluation of cell tracking algorithms has been established under the auspices of the IEEE International Symposium on Biomedical Imaging 2013 Cell Tracking Challenge. In this article, we present the logistics, datasets, methods and results of the challenge and lay down the principles for future uses of this benchmark.

Results: The main contributions of the challenge include the creation of a comprehensive video dataset repository and the definition of objective measures for comparison and ranking of the algorithms. With this benchmark, six algorithms covering a variety of segmentation and tracking paradigms have been compared and ranked based on their performance on both synthetic and real datasets. Given the diversity of the datasets, we do not declare a single winner of the challenge. Instead, we present and discuss the results for each individual dataset separately.

Availability and implementation: The challenge Web site (http://www.codesolorzano.com/celltrackingchallenge) provides access to the training and competition datasets, along with the ground truth of the training videos. It also provides access to Windows and Linux executable files of the evaluation software and most of the algorithms that competed in the challenge.

Contact:codesolorzano@unav.es

Supplementary information:Supplementary data are available at Bioinformatics online.

Optimal multiresolution 3D level-set method for liver segmentation incorporating local curvature constraints

Advanced liver surgery requires a precise pre-operative planning, where liver segmentation and remnant liver volume are key elements to avoid post-operative liver failure. In that context, level-set algorithms have achieved better results than others, especially with altered liver parenchyma or in cases with previous surgery. In order to improve functional liver parenchyma volume measurements, in this work we propose two strategies to enhance previous level-set algorithms: an optimal multi-resolution strategy with fine details correction and adaptive curvature, as well as an additional semiautomatic step imposing local curvature constraints. Results show more accurate segmentations, especially in elongated structures, detecting internal lesions and avoiding leakages to close structures.