SAF: Smart Aggregation Framework for Revealing Atoms Importance Rank and Improving Prediction Rates in Drug Discovery.

Machine learning, and representation learning in particular, has the potential to facilitate drug discovery by screening a large chemical space in silico. A successful approach for representing molecules is to treat them as graphs and utilize graph neural networks. One of the key limitations of such methods is the necessity to represent compounds with different numbers of atoms, which requires aggregating the atom's information. Common aggregation operators, such as averaging, result in a loss of information at the atom level. In this work, we propose a novel aggregating approach where each atom is weighted nonlinearly using the Boltzmann distribution with a hyperparameter analogous to temperature. We show that using this weighted aggregation improves the ability of the gold standard message-passing neural network to predict antibiotic activity. Moreover, by changing the temperature hyperparameter, our approach can reveal the atoms that are important for activity prediction in a smooth and consistent way, thus providing a novel regulated attention mechanism for graph neural networks. We further validate our method by showing that it recapitulates the functional group in ß-lactam antibiotics. The ability of our approach to rank the atoms' importance for a desired function can be used within any graph neural network to provide interpretability of the results and predictions at the node level.

DEPAS: De-novo Pathology Semantic Masks using a Generative Model.

The integration of artificial intelligence (AI) into digital pathology has the potential to automate and improve various tasks, such as image analysis and diagnostic decision-making. Yet, the inherent variability of tissues, together with the need for image labeling, lead to biased datasets that limit the generalizability of algorithms trained on them. One of the emerging solutions for this challenge is synthetic histological images. Debiasing real datasets require not only generating photorealistic images but also the ability to control the cellular features within them. A common approach is to use generative methods that perform image translation between semantic masks that reflect prior knowledge of the tissue and a histological image. However, unlike other image domains, the complex structure of the tissue prevents a simple creation of histology semantic masks that are required as input to the image translation model, while semantic masks extracted from real images reduce the process's scalability. In this work, we introduce a scalable generative model, coined as DEPAS (De-novo Pathology Semantic Masks), that captures tissue structure and generates high-resolution semantic masks with state-of-the-art quality. We demonstrate the ability of DEPAS to generate realistic semantic maps of tissue for three types of organs: skin, prostate, and lung. Moreover, we show that these masks can be processed using a generative image translation model to produce photorealistic histology images of two types of cancer with two different types of staining techniques. Finally, we harness DEPAS to generate multi-label semantic masks that capture different cell types distributions and use them to produce histological images with on-demand cellular features. Overall, our work provides a state-of-the-art solution for the challenging task of generating synthetic histological images while controlling their semantic information in a scalable way.

Between Generating Noise and Generating Images: Noise in the Correct Frequency Improves the Quality of Synthetic Histopathology Images for Digital Pathology.

Artificial intelligence and machine learning techniques have the promise to revolutionize the field of digital pathology. However, these models demand considerable amounts of data, while the availability of unbiased training data is limited. Synthetic images can augment existing datasets, to improve and validate AI algorithms. Yet, controlling the exact distribution of cellular features within them is still challenging. One of the solutions is harnessing conditional generative adversarial networks that take a semantic mask as an input rather than a random noise. Unlike other domains, outlining the exact cellular structure of tissues is hard, and most of the input masks depict regions of cell types. This is also the case for non-small cell lung cancer, the most common type of lung cancer. Deciding whether a patient would receive immunotherapy depends on quantifying regions of stained cells. However, using polygon-based masks introduce inherent artifacts within the synthetic images - due to the mismatch between the polygon size and the single-cell size. In this work, we show that introducing random single-pixel noise with the appropriate spatial frequency into a polygon semantic mask can dramatically improve the quality of the synthetic images. We used our platform to generate synthetic images of immunohistochemistry-treated lung biopsies. We test the quality of the images using a three-fold validation procedure. First, we show that adding the appropriate noise frequency yields 87% of the similarity metrics improvement that is obtained by adding the actual single-cell features. Second, we show that the synthetic images pass the Turing test. Finally, we show that adding these synthetic images to the train set improves AI performance in terms of PD-L1 semantic segmentation performances. Our work suggests a simple and powerful approach for generating synthetic data on demand to unbias limited datasets to improve the algorithms' accuracy and validate their robustness.

Single-cell analysis of the aged ovarian immune system reveals a shift towards adaptive immunity and attenuated cell function.

The immune system plays a major role in maintaining many physiological processes in the reproductive system. However, a complete characterization of the immune milieu in the ovary, and particularly how it is affected by female aging, is still lacking. Here, we utilize single-cell RNA sequencing and flow cytometry to construct the complete description of the murine ovarian immune system. We show that the composition of the immune cells undergoes an extensive shift with age towards adaptive immunity. We analyze the effect of aging on gene expression and chemokine and cytokine networks and show an overall decreased expression of inflammatory mediators together with an increased expression of senescent cells recognition receptors. Our results suggest that the fertile female's ovarian immune aging differs from the suggested female post-menopause inflammaging as it copes with the inflammatory stimulations during repeated cycles and the increasing need for clearance of accumulating atretic follicles.

Harnessing artificial intelligence to infer novel spatial biomarkers for the diagnosis of eosinophilic esophagitis.

Eosinophilic esophagitis (EoE) is a chronic allergic inflammatory condition of the esophagus associated with elevated esophageal eosinophils. Second only to gastroesophageal reflux disease, EoE is one of the leading causes of chronic refractory dysphagia in adults and children. EoE is a clinicopathologic disorder and the histological portion of the diagnosis requires enumerating the density of esophageal eosinophils in esophageal biopsies, and evaluating additional features such as basal zone hyperplasia is helpful. However, this task requires time-consuming, somewhat subjective manual analysis, thus reducing the ability to process the complex tissue structure and infer its relationship with the patient's clinical status. Previous artificial intelligence (AI) approaches that aimed to improve histology-based diagnosis focused on recapitulating identification and quantification of the area of maximal eosinophil density, the gold standard manual metric for determining EoE disease activity. However, this metric does not account for the distribution of eosinophils or other histological features, over the whole slide image. Here, we developed an artificial intelligence platform that infers local and spatial biomarkers based on semantic segmentation of intact eosinophils and basal zone distributions. Besides the maximal density of eosinophils [referred to as Peak Eosinophil Count (PEC)] and a maximal basal zone fraction, we identify the value of two additional metrics that reflect the distribution of eosinophils and basal zone fractions. This approach enables a decision support system that predicts EoE activity and potentially classifies the histological severity of EoE patients. We utilized a cohort that includes 1,066 biopsy slides from 400 subjects to validate the system's performance and achieved a histological severity classification accuracy of 86.70%, sensitivity of 84.50%, and specificity of 90.09%. Our approach highlights the importance of systematically analyzing the distribution of biopsy features over the entire slide and paves the way toward a personalized decision support system that will assist not only in counting cells but can also potentially improve diagnosis and provide treatment prediction.

Density-dependent effects are the main determinants of variation in growth dynamics between closely related bacterial strains.

Although closely related, bacterial strains from the same species show significant diversity in their growth and death dynamics. Yet, our understanding of the relationship between the kinetic parameters that dictate these dynamics is still lacking. Here, we measured the growth and death dynamics of 11 strains of Escherichia coli originating from different hosts and show that the growth patterns are clustered into three major classes with typical growth rates, maximal fold change, and death rates. To infer the underlying phenotypic parameters that govern the dynamics, we developed a phenomenological mathematical model that accounts not only for growth rate and its dependence on resource availability, but also for death rates and density-dependent growth inhibition. We show that density-dependent growth is essential for capturing the variability in growth dynamics between the strains. Indeed, the main parameter determining the dynamics is the typical density at which they slow down their growth, rather than the maximal growth rate or death rate. Moreover, we show that the phenotypic landscape resides within a two-dimensional plane spanned by resource utilization efficiency, death rate, and density-dependent growth inhibition. In this phenotypic plane, we identify three clusters that correspond to the growth pattern classes. Overall, our results reveal the tradeoffs between growth parameters that constrain bacterial adaptation.

A Deep Multi-Label Segmentation Network For Eosinophilic Esophagitis Whole Slide Biopsy Diagnostics.

Eosinophilic esophagitis (EoE) is an allergic inflammatory condition of the esophagus associated with elevated numbers of eosinophils. Disease diagnosis and monitoring require determining the concentration of eosinophils in esophageal biopsies, a time-consuming, tedious and somewhat subjective task currently performed by pathologists. Here, we developed a machine learning pipeline to identify, quantitate and diagnose EoE patients' at the whole slide image level. We propose a platform that combines multi-label segmentation deep network decision support system with dynamics convolution that is able to process whole biopsy slide. Our network is able to segment both intact and not-intact eosinophils with a mean intersection over union (mIoU) of 0.93. This segmentation enables the local quantification of intact eosinophils with a mean absolute error of 0.611 eosinophils. We examined a cohort of 1066 whole slide images from 400 patients derived from multiple institutions. Using this set, our model achieved a global accuracy of 94.75%, sensitivity of 94.13%, and specificity of 95.25% in reporting EoE disease activity. Our work provides state-of-the-art performances on the largest EoE cohort to date, and successfully addresses two of the main challenges in EoE diagnostics and digital pathology, the need to detect several types of small features simultaneously, and the ability to analyze whole slides efficiently. Our results pave the way for an automated diagnosis of EoE and can be utilized for other conditions with similar challenges.

Machine Learning Approach for Biopsy-Based Identification of Eosinophilic Esophagitis Reveals Importance of Global features.

GOAL: Eosinophilic esophagitis (EoE) is an allergic inflammatory condition characterized by eosinophil accumulation in the esophageal mucosa. EoE diagnosis includes a manual assessment of eosinophil levels in mucosal biopsies-a time-consuming, laborious task that is difficult to standardize. One of the main challenges in automating this process, like many other biopsy-based diagnostics, is detecting features that are small relative to the size of the biopsy. RESULTS: In this work, we utilized hematoxylin- and eosin-stained slides from esophageal biopsies from patients with active EoE and control subjects to develop a platform based on a deep convolutional neural network (DCNN) that can classify esophageal biopsies with an accuracy of 85%, sensitivity of 82.5%, and specificity of 87%. Moreover, by combining several downscaling and cropping strategies, we show that some of the features contributing to the correct classification are global rather than specific, local features. CONCLUSIONS: We report the ability of artificial intelligence to identify EoE using computer vision analysis of esophageal biopsy slides. Further, the DCNN features associated with EoE are based on not only local eosinophils but also global histologic changes. Our approach can be used for other conditions that rely on biopsy-based histologic diagnostics.

Discrete limbal epithelial stem cell populations mediate corneal homeostasis and wound healing.

The accessibility and transparency of the cornea permit robust stem cell labeling and in vivo cell fate mapping. Limbal epithelial stem cells (LSCs) that renew the cornea are traditionally viewed as rare, slow-cycling cells that follow deterministic rules dictating their self-renewal or differentiation. Here, we combined single-cell RNA sequencing and advanced quantitative lineage tracing for in-depth analysis of the murine limbal epithelium. These analysis revealed the co-existence of two LSC populations localized in separate and well-defined sub-compartments, termed the "outer" and "inner" limbus. The primitive population of quiescent outer LSCs participates in wound healing and boundary formation, and these cells are regulated by T cells, which serve as a niche. In contrast, the inner peri-corneal limbus hosts active LSCs that maintain corneal epithelial homeostasis. Quantitative analyses suggest that LSC populations are abundant, following stochastic rules and neutral drift dynamics. Together these results demonstrate that discrete LSC populations mediate corneal homeostasis and regeneration.

Spatial correlations constrain cellular lifespan and pattern formation in corneal epithelium homeostasis.

Homeostasis in adult tissues relies on the replication dynamics of stem cells, their progenitors and the spatial balance between them. This spatial and kinetic coordination is crucial to the successful maintenance of tissue size and its replenishment with new cells. However, our understanding of the role of cellular replicative lifespan and spatial correlation between cells in shaping tissue integrity is still lacking. We developed a mathematical model for the stochastic spatial dynamics that underlie the rejuvenation of corneal epithelium. Our model takes into account different spatial correlations between cell replication and cell removal. We derive the tradeoffs between replicative lifespan, spatial correlation length, and tissue rejuvenation dynamics. We determine the conditions that allow homeostasis and are consistent with biological timescales, pattern formation, and mutants phenotypes. Our results can be extended to any cellular system in which spatial homeostasis is maintained through cell replication.

Technological Developments and Strategic Management for Overcoming the COVID-19 Challenge within the Hospital Setting in Israel.

The coronavirus disease 2019 (COVID-19) pandemic has remarkably challenged health care organizations and societies. A key strategy for confronting the disease implications on individuals and communities was based on harnessing multidisciplinary efforts to develop technologies for mitigating the disease spread and its deleterious clinical implications. One of the main challenging characteristics of COVID-19 is the provision of medical care to patients with a highly infective disease mandating the use of isolation measures. Such care is complicated by the need for complex critical care, dynamic treatment guidelines, and a vague knowledge regarding the disease's pathophysiology. A second key component of this challenge was the overwhelming surge in patient burden and the relative lack of trained staff and medical equipment which required rapid re-organization of large systems and augmenting health care efficiencies to unprecedented levels. In contrast to the risk management strategies employed to mitigate other serious threats and the billions of dollars that are invested in reducing these risks annually by governments around the world, no such preparation has been shown to be of effect during the current COVID-19 pandemic. Unmet needs were identified within the newly opened COVID-19 departments together with the urgent need for reliable information for effective decision-making at the state level.This review article describes the early research and development response in Israel under the scope of in-hospital patient care, such as non-contact sensing of patients' vital signs, and how it could potentially be weaved into a practical big picture at the hospital or national level using a strategic management system. At this stage, some of the described technologies are still in developmental or clinical evidence generation phases with respect to COVID-19 settings. While waiting for future publications describing the results of the ongoing evidence generation efforts, one should be aware of this trend as these emerging tools have the potential to further benefit patients as well as caregivers and health care systems beyond the scope of the current pandemic as well as confronting future surges in the number of cases.

How to design an optimal sensor network for the unfolded protein response.

Cellular protein homeostasis requires continuous monitoring of stress in the endoplasmic reticulum (ER). Stress-detection networks control protein homeostasis by mitigating the deleterious effects of protein accumulation, such as aggregation and misfolding, with precise modulation of chaperone production. Here, we develop a coarse model of the unfolded protein response in yeast and use multi-objective optimization to determine which sensing and activation strategies optimally balance the trade-off between unfolded protein accumulation and chaperone production. By comparing a stress-sensing mechanism that responds directly to the level of unfolded protein in the ER to a mechanism that is negatively regulated by unbound chaperones, we show that chaperone-mediated sensors are more efficient than sensors that detect unfolded proteins directly. This results from the chaperone-mediated sensor having separate thresholds for activation and deactivation. Finally, we demonstrate that a sensor responsive to both unfolded protein and unbound chaperone does not further optimize homeostatic control. Our results suggest a strategy for designing stress sensors and may explain why BiP-mitigated ER stress-sensing networks have evolved.

Nanoliter Cell Culture Array with Tunable Chemical Gradients.

A multitude of cell screening assays for diagnostic and research applications rely on quantitative measurements of a sample in the presence of different reagent concentrations. Standard methods rely on microtiter plates of varying well density, which provide simple and standardized sample addressability. However, testing hundreds of chemical dilutions requires complex automation, and typical well volumes of microtiter plates are incompatible with the analysis of a small number of cells. Here, we present a microfluidic device for creating a high-resolution chemical gradient spanning 200 nanoliter wells. Using air-based shearing, we show that the individual wells can be compartmentalized without altering the concentration gradient, resulting in a large set of isolated nanoliter cell culture wells. We provide an analytical and numerical model for predicting the concentration within each culture chamber and validate it against experimental results. We apply our system for the investigation of yeast cell metabolic gene regulation in the presence of different ratios of galactose/glucose concentrations and successfully resolve the nutrient threshold at which the cells activate the galactose pathway.

Achieving global perfect homeostasis through transporter regulation.

Nutrient homeostasis-the maintenance of relatively constant internal nutrient concentrations in fluctuating external environments-is essential to the survival of most organisms. Transcriptional regulation of plasma membrane transporters by internal nutrient concentrations is typically assumed to be the main mechanism by which homeostasis is achieved. While this mechanism is homeostatic we show that it does not achieve global perfect homeostasis-a condition where internal nutrient concentrations are completely independent of external nutrient concentrations for all external nutrient concentrations. We show that the criterion for global perfect homeostasis is that transporter levels must be inversely proportional to net nutrient flux into the cell and that downregulation of active transporters (activity-dependent regulation) is a simple and biologically plausible mechanism that meets this criterion. Activity-dependent transporter regulation creates a trade-off between robustness and efficiency, i.e., the system's ability to withstand perturbation in external nutrients and the transporter production rate needed to maintain homeostasis. Additionally, we show that a system that utilizes both activity-dependent transporter downregulation and regulation of transporter synthesis by internal nutrient levels can create a system that mitigates the shortcomings of each of the individual mechanisms. This analysis highlights the utility of activity-dependent regulation in achieving homeostasis and calls for a re-examination of the mechanisms of regulation of other homeostatic systems.

Comment on "Ribosome utilizes the minimum free energy changes to achieve the highest decoding rate and fidelity".

We examined [Y. Savir and T. Tlusty, Cell 153, 471 (2013)10.1016/j.cell.2013.03.032] the decoding performance of tRNA by the ribosome. For this purpose, we specified the kinetics of tRNA decoding and the corresponding energy landscape, from which we calculated the steady-state decoding rate R_{C}. Following our work, Xie reexamined [P. Xie, Phys. Rev. E 92, 022716 (2015)10.1103/PhysRevE.92.022716] the energy landscape of tRNA decoding. His analysis relies on an alternative expression for R_{C}, while claiming that the expression we use is missing some terms. In this Comment we rederive in detail our expression for the steady-state decoding rate R_{C}, show they hold, explain why the alternative expression for R_{C} is inaccurate, and discuss the underlying intuition.

Identifying Metabolic Subpopulations from Population Level Mass Spectrometry.

Metabolism underlies many important cellular decisions, such as the decisions to proliferate and differentiate, and defects in metabolic signaling can lead to disease and aging. In addition, metabolic heterogeneity can have biological consequences, such as differences in outcomes and drug susceptibilities in cancer and antibiotic treatments. Many approaches exist for characterizing the metabolic state of a population of cells, but technologies for measuring metabolism at the single cell level are in the preliminary stages and are limited. Here, we describe novel analysis methodologies that can be applied to established experimental methods to measure metabolic variability within a population. We use mass spectrometry to analyze amino acid composition in cells grown in a mixture of (12)C- and (13)C-labeled sugars; these measurements allow us to quantify the variability in sugar usage and thereby infer information about the behavior of cells within the population. The methodologies described here can be applied to a large range of metabolites and macromolecules and therefore have the potential for broad applications.

Competitive inhibition can linearize dose-response and generate a linear rectifier.

Many biological responses require a dynamic range that is larger than standard bi-molecular interactions allow, yet the also ability to remain off at low input. Here we mathematically show that an enzyme reaction system involving a combination of competitive inhibition, conservation of the total level of substrate and inhibitor, and positive feedback can behave like a linear rectifier-that is, a network motif with an input-output relationship that is linearly sensitive to substrate above a threshold but unresponsive below the threshold. We propose that the evolutionarily conserved yeast SAGA histone acetylation complex may possess the proper physiological response characteristics and molecular interactions needed to perform as a linear rectifier, and we suggest potential experiments to test this hypothesis. One implication of this work is that linear responses and linear rectifiers might be easier to evolve or synthetically construct than is currently appreciated.

Natural variation in preparation for nutrient depletion reveals a cost-benefit tradeoff.

Maximizing growth and survival in the face of a complex, time-varying environment is a common problem for single-celled organisms in the wild. When offered two different sugars as carbon sources, microorganisms first consume the preferred sugar, then undergo a transient growth delay, the "diauxic lag," while inducing genes to metabolize the less preferred sugar. This delay is commonly assumed to be an inevitable consequence of selection to maximize use of the preferred sugar. Contrary to this view, we found that many natural isolates of Saccharomyces cerevisiae display short or nonexistent diauxic lags when grown in mixtures of glucose (preferred) and galactose. These strains induce galactose utilization (GAL) genes hours before glucose exhaustion, thereby "preparing" for the transition from glucose to galactose metabolism. The extent of preparation varies across strains, and seems to be determined by the steady-state response of GAL genes to mixtures of glucose and galactose rather than by induction kinetics. Although early GAL gene induction gives strains a competitive advantage once glucose runs out, it comes at a cost while glucose is still present. Costs and benefits correlate with the degree of preparation: strains with higher expression of GAL genes prior to glucose exhaustion experience a larger upfront growth cost but also a shorter diauxic lag. Our results show that classical diauxic growth is only one extreme on a continuum of growth strategies constrained by a cost-benefit tradeoff. This type of continuum is likely to be common in nature, as similar tradeoffs can arise whenever cells evolve to use mixtures of nutrients.

Galactose metabolic genes in yeast respond to a ratio of galactose and glucose.

Natural environments are filled with multiple, often competing, signals. In contrast, biological systems are often studied in "well-controlled" environments where only a single input is varied, potentially missing important interactions between signals. Catabolite repression of galactose by glucose is one of the best-studied eukaryotic signal integration systems. In this system, it is believed that galactose metabolic (GAL) genes are induced only when glucose levels drop below a threshold. In contrast, we show that GAL gene induction occurs at a constant external galactose:glucose ratio across a wide range of sugar concentrations. We systematically perturbed the components of the canonical galactose/glucose signaling pathways and found that these components do not account for ratio sensing. Instead we provide evidence that ratio sensing occurs upstream of the canonical signaling pathway and results from the competitive binding of the two sugars to hexose transporters. We show that a mutant that behaves as the classical model expects (i.e., cannot use galactose above a glucose threshold) has a fitness disadvantage compared with wild type. A number of common biological signaling motifs can give rise to ratio sensing, typically through negative interactions between opposing signaling molecules. We therefore suspect that this previously unidentified nutrient sensing paradigm may be common and overlooked in biology.