Deep neural models for color classification and color constancy

Color constancy is our ability to perceive constant colors across varying illuminations. Here, we trained deep neural networks to be color constant and evaluated their performance with varying cues. Inputs to the networks consisted of two-dimensional images of simulated cone excitations derived from three-dimensional (3D) rendered scenes of 2,115 different 3D shapes, with spectral reflectances of 1,600 different Munsell chips, illuminated under 278 different natural illuminations. The models were trained to classify the reflectance of the objects. Testing was done with four new illuminations with equally spaced CIEL*a*b* chromaticities, two along the daylight locus and two orthogonal to it. High levels of color constancy were achieved with different deep neural networks, and constancy was higher along the daylight locus. When gradually removing cues from the scene, constancy decreased. Both ResNets and classical ConvNets of varying degrees of complexity performed well. However, DeepCC, our simplest sequential convolutional network, represented colors along the three color dimensions of human color vision, while ResNets showed a more complex representation.     

Free Zinc as a Predictive Marker for COVID-19 Mortality Risk

Free zinc is considered to be the exchangeable and biological active form of zinc in serum, and is discussed to be a suitable biomarker for alterations in body zinc homeostasis and related diseases. Given that coronavirus disease 2019 (COVID-19) is characterized by a marked decrease in total serum zinc, and clinical data indicate that zinc status impacts the susceptibility and severity of the infection, we hypothesized that free zinc in serum might be altered in response to SARS-CoV-2 infection and may reflect disease severity. To test this hypothesis, free zinc concentrations in serum samples of survivors and nonsurvivors of COVID-19 were analyzed by fluorometric microassay. Similar to the reported total serum zinc deficit measured by total reflection X-ray fluorescence, free serum zinc in COVID-19 patients was considerably lower than that in control subjects, and surviving patients displayed significantly higher levels of free zinc than those of nonsurvivors (mean ± SD; 0.4 ± 0.2 nM vs. 0.2 ± 0.1 nM; p = 0.0004). In contrast to recovering total zinc concentrations (r = 0.706, p < 0.001) or the declining copper–zinc ratio (r = −0.646; p < 0.001), free zinc concentrations remained unaltered with time in COVID-19 nonsurvivors. Free serum zinc concentrations were particularly low in male as compared to female patients (mean ± SD; 0.4 ± 0.2 nM vs. 0.2 ± 0.1 nM; p = 0.0003). This is of particular interest, as the male sex is described as a risk factor for severe COVID-19. Overall, results indicate that depressed free serum zinc levels are associated with increased risk of death in COVID-19, suggesting that free zinc may serve as a novel prognostic marker for the severity and course of COVID-19.     

The Impact of COVID-19 on Routine Medical Care and Cancer Screening

BackgroundCOVID-19 restrictions and fear dramatically changed the use of medical care. Understanding the magnitude of cancelled and postponed appointments and associated factors can help identify approaches to mitigate unmet need.ObjectiveTo determine the proportion of medical visits cancelled or postponed and for whom. We hypothesized that adults with serious medical conditions and those with higher anxiety, depressive symptoms, and avoidance-oriented coping would have more cancellations/postponements.DesignFour nationally representative cross-sectional surveys conducted online in May, July, October, and December 2020.Participants59,747 US adults who completed 15-min online surveys. 69% cooperation rate.MeasuresPhysical and mental health visits and cancer screening cancelled or postponed over prior 2 months. Plan to cancel or postpone visits over the next 2 months. Relationship with demographics, medical conditions, local COVID-19 death rate, anxiety, depressive symptoms, coping, intolerance of uncertainty, and perceived COVID-19 risk.Key ResultsOf the 58% (N = 34,868) with a medical appointment during the 2 months before the survey, 64% had an appointment cancelled or postponed in May, decreasing to 37% in December. Of the 41% of respondents with scheduled cancer screening, 20% cancelled/postponed, which was stable May to December. People with more medical conditions were more likely to cancel or postpone medical visits (OR 1.19 per condition, 95% CI 1.16, 1.22) and cancer screening (OR 1.20, 95% CI 1.15, 1.24). Race, ethnicity, and income had weak associations with cancelled/postponed visits, local death rate was unrelated, but anxiety and depressive symptoms were strongly related to cancellations, and this grew between May and December.ConclusionsCancelled medical care and cancer screening were more common among persons with medical conditions, anxiety and depression, even after accounting for COVID-19 deaths. Outreach and support to ensure that patients are not avoiding needed care due to anxiety, depression and inaccurate perceptions of risk will be important.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11606-021-07254-x.KEY WORDS: COVID-19, missed medical appointments, cancer screening     

A protease-mediated switch regulates the growth of magnetosome organelles in Magnetospirillum magneticum

Magnetosomes are lipid-bound organelles that direct the biomineralization of magnetic nanoparticles in magnetotactic bacteria. Magnetosome membranes are not uniform in size and can grow in a biomineralization-dependent manner. However, the underlying mechanisms of magnetosome membrane growth regulation remain unclear. Using cryoelectron tomography, we systematically examined mutants with defects at various stages of magnetosome formation to identify factors involved in controlling membrane growth. We found that a conserved serine protease, MamE, plays a key role in magnetosome membrane growth regulation. When the protease activity of MamE is disrupted, magnetosome membrane growth is restricted, which, in turn, limits the size of the magnetite particles. Consistent with this finding, the upstream regulators of MamE protease activity, MamO and MamM, are also required for magnetosome membrane growth. We then used a combination of candidate and comparative proteomics approaches to identify Mms6 and MamD as two MamE substrates. Mms6 does not appear to participate in magnetosome membrane growth. However, in the absence of MamD, magnetosome membranes grow to a larger size than the wild type. Furthermore, when the cleavage of MamD by MamE protease is blocked, magnetosome membrane growth and biomineralization are severely inhibited, phenocopying the MamE protease-inactive mutant. We therefore propose that the growth of magnetosome membranes is controlled by a protease-mediated switch through processing of MamD. Overall, our work shows that, like many eukaryotic systems, bacteria control the growth and size of biominerals by manipulating the physical properties of intracellular organelles.

Biomineralization is a common phenomenon across the tree of life; one type is a biologically controlled mineral production process that is often initiated within intracellular membrane-bound organelles or vesicle-like structures (1, 2). For instance, matrix vesicles serve as initial sites for mineral formation in the growth plate and most other vertebrate mineralization tissues (3). Vesicles also play a central role in the formation of calcitic spicules in sea urchins (4), extracellular calcitic plates in marine coccolithophores (5), and the silica-based cell walls of diatoms (6). Compartmentalization within a membrane is believed to provide an isolated microenvironment and a template for efficient nucleation, growth, and shaping of minerals.In contrast to the multiple examples of eukaryotic biomineralization noted here, little is known regarding the diversity and dynamics of bacterial biomineralization at the molecular and cellular level. Production of magnetic minerals within magnetosome organelles of magnetotactic bacteria (MTB) stands as one of the best-studied examples of biomineralization in bacteria. MTB are a diverse group of gram-negative bacteria often found near the oxic-anoxic transition zone of aquatic environments (7). Magnetic nanoparticles (magnetite or greigite) mineralized by MTB are generally 35 to 120 nm in length, a size range that yields a single, stable magnetic moment (8). Magnetosomes are typically arranged into one or multiple chains that function as a complete magnetic unit, enabling MTB to navigate along geomagnetic field lines and efficiently find the oxic-anoxic transition zone in a process termed magneto-aerotaxis (9).Biomineralization compartments generally contain a specific cohort of proteins that play critical roles during organelle formation and mineralization. Proteins involved in magnetosome biogenesis are normally encoded from a genomic region called the magnetosome gene island (MAI) (SI Appendix, Fig. S1A). Many magnetosome‐associated membrane (Mam) and magnetic particle membrane‐specific (Mms) proteins are associated with magnetosomes (10–12). The genes encoding the Mam and Mms proteins are organized into four clusters (mamAB, mamGFDC, mms6, and mamXY) in the model Magnetosprillum species and are necessary and sufficient for magnetosome formation (13–17) (Fig. 1A). Analyses of deletion mutants have been used to assign roles for individual genes in one of four distinct stages of magnetosome biogenesis in the model organism Magnetospirillum magneticum AMB-1 (AMB-1): 1) empty membrane invagination (mamI, -L, -Q, and -B), 2) chain alignment (mamK, -J, and -Y), 3) crystal nucleation (mamM, -N, and -O), and 4) crystal maturation (other genes within the four clusters) (13, 14, 18) (Fig. 1A).Open in a separate windowFig. 1.The essential genes and the process of magnetosome production. (A) Schematic depicting the four key magnetosome gene clusters of AMB-1. A total of 10 genes were tested in this study for magnetosome membrane growth regulation: genes involved in crystal initiation (mamM, -N, and -O) are marked in orange, and genes involved in crystal maturation (mamE, -P, -A, -S, -T, -D, and mms6) are marked in blue. Based on previous work, mmsF is known to not be involved in magnetosome membrane growth (8). (B) Model of the biomineralization-dependent magnetosome membrane growth based on Cornejo et al. 2016 (19). OM, outer membrane. IM, inner membrane.The resulting stepwise model outlines a set of processes that are seemingly distinct from one another. However, examination of the dynamics of magnetosome formation has revealed that magnetosome membrane growth is closely linked to the progression of biomineralization. Within a given AMB-1 cell, the magnetosome chain consists of some empty magnetosome membranes (EMMs) as well as the crystal-containing magnetosome membranes (CMMs) that provide the dipole moment necessary for orientation in magnetic fields. Cornejo et al. showed that at steady state, the diameter of the magnetosome lumen ranges from 20 to 80 nm (volume of about 4,189 to 268,083 nm³), yet no EMMs grow beyond 55 nm (volume of about 65,450 nm³) (19). Accordingly, when biomineralization is disrupted by limiting iron availability, only EMMs are produced and their growth stalls at about 55 nm, implying the existence of a checkpoint for membrane growth (Fig. 1B). Upon iron addition, membranes that have initiated biomineralization (CMMs) grow larger than this limit, implying that active biomineralization is needed for further membrane growth (19) leading to a linear relationship between the size of the growing crystals and the surrounding membranes (Fig. 1B). One possible explanation for these observations is that the growing mineral pushes against the membrane and drives its expansion. However, a mutant missing MmsF, a late-stage biomineralization protein, makes small magnetite crystals and still produces membranes as large as the wild-type (WT) parent in a biomineralization-dependent manner (19). These observations imply that magnetosome membrane growth is tightly regulated to create an optimal environment for crystal nucleation, which triggers the second membrane growth stage for crystal maturation (19).The discovery of biomineralization-dependent magnetosome membrane growth provides a lens to examine the function of magnetosome proteins. One hypothesis holds that regulated growth of the magnetosome membrane allows for proper accumulation of iron to high concentrations to initiate nucleation and growth of magnetic particles. Thus, factors known to influence the growth and geometry of magnetite crystals may actually do so by regulating the physical properties of the magnetosome membrane. Here, we explored this possibility by using whole cell cryoelectron tomography (cryo-ET) to directly measure the sizes of magnetosome membranes in a series of mutants with known defects in crystal production. MamE belongs to a highly conserved high temperature requirement A (HtrA) family of trypsin-like serine proteases, whose activity is required for crystal maturation with an unknown mechanism (20). Here, we find that the catalytic activity of MamE plays a central role in the progression of magnetosome membrane growth. MamE proteolytically processes itself and two other biomineralization factors, MamO and MamP (21). MamO is required for MamE protease activation (22), and we find it acts as an upstream regulator of MamE for magnetosome membrane growth. MamP is not involved in magnetosome membrane growth. We also identified MamD, a protein that binds tightly to magnetite and was previously thought to promote crystal maturation (23, 24), as a direct substrate of MamE and showed that MamD is in fact a negative regulator of biomineralization. Our results indicate that MamE activates membrane remodeling by relieving MamD’s inhibition on the size of the magnetosome lumen and demonstrate how spatial restructuring of an organelle can regulate its biochemical output.     

Assessment of Left Ventricular Viscoelastic Components Based on Ventricular Harmonic Behavior

Background: Assessment of left ventricular (LV) function with an emphasis on contractility has been a challenge in cardiac mechanics during the recent decades. The LV function is usually described by the LV pressure-volume (P-V) relationship. Based on this relationship, the ratio of instantaneous pressure to instantaneous volume is an index for LV chamber stiffness. The standard P-V diagrams are easy to interpret but difficult to obtain and require invasive instrumentation for measuring the corresponding volume and pressure data. In the present study, we introduce a technique that can estimate viscoelastic properties, not only the elastic component but also the viscous properties of the LV based on oscillatory behavior of the ventricular chamber and it can be applied non-invasively as well. Materials and Methods: The estimation technique is based on modeling the actual long axis displacement of the mitral annulus plane toward the cardiac base as a linear damped oscillator with time-varying coefficients. Elastic deformations resulting from the changes in the ventricular mechanical properties of myocardium are represented as a time-varying spring while the viscous components of the model include a time-varying viscous damper, representing relaxation and the frictional energy loss. To measure the left ventricular axial displacement ten healthy sheep underwent left thoracotomy and sonomicrometry transducers were implanted at the apex and base of the LV. The time-varying parameters of the model were estimated by a standard Recursive Linear Least Squares (RLLS) technique. Results: LV stiffness at end-systole and end-diastole was in the range of 61.86–136 dyne/g.cm and 1.25–21.02 dyne/g.cm, respectively. Univariate linear regression was performed to verify the agreement between the estimated parameters, and the measured values of stiffness. The averaged magnitude of the stiffness and damping coefficients during a complete cardiac cycle were estimated as 58.63±12.8 dyne/g.cm and 0 dyne.s/g.cm, respectively. Conclusion: The results for the estimated elastic coefficients are consistent with the ones obtained from force-displacement diagram. The trend of change in the estimated parameters is also in harmony with the previous studies done using P-V diagram. The only input used in this model is the long axis displacement of the annulus plane, which can also be obtained non-invasively using tissue Doppler or MR imaging.     

A mechanism for sensory re-weighting in postural control

A key finding of human balance experiments has been that the integration of sensory information utilized for postural control appears to be dynamically regulated to adapt to changing environmental conditions and the available sensory information, a process referred to as “sensory re-weighting.” We propose a postural control model that includes automatic sensory re-weighting. This model is an adaptation of a previously reported model of sensory feedback that included manual sensory re-weighting. The new model achieves sensory re-weighting that is physiologically plausible and readily implemented. Model simulations are compared to previously reported experimental results to demonstrate the automated sensory re-weighting strategy of the modified model. On the whole, the postural sway time series generated by the model with automatic sensory re-weighting show good agreement with experimental data, and are capable of producing patterns similar to those observed experimentally.