The implementation of clinical-decision support algorithms for medical imaging faces challenges with reliability and interpretability. Here, we establish a diagnostic tool based on a deep-learning framework for the screening of patients with common treatable blinding retinal diseases. Our framework utilizes transfer learning, which trains a neural network with a fraction of the data of conventional approaches. Applying this approach to a dataset of optical coherence tomography images, we demonstrate performance comparable to that of human experts in classifying age-related macular degeneration and diabetic macular edema. We also provide a more transparent and interpretable diagnosis by highlighting the regions recognized by the neural network. We further demonstrate the general applicability of our AI system for diagnosis of pediatric pneumonia using chest X-ray images. This tool may ultimately aid in expediting the diagnosis and referral of these treatable conditions, thereby facilitating earlier treatment, resulting in improved clinical outcomes. VIDEO ABSTRACT.

Determinacy is an agronomically important trait associated with the domestication in soybean (Glycine max). Most soybean cultivars are classifiable into indeterminate and determinate growth habit, whereas Glycine soja, the wild progenitor of soybean, is indeterminate. Indeterminate (Dt1/Dt1) and determinate (dt1/dt1) genotypes, when mated, produce progeny that segregate in a monogenic pattern. Here, we show evidence that Dt1 is a homolog (designated as GmTfl1) of Arabidopsis terminal flower 1 (TFL1), a regulatory gene encoding a signaling protein of shoot meristems. The transition from indeterminate to determinate phenotypes in soybean is associated with independent human selections of four distinct singlenucleotide substitutions in the GmTfl1 gene, each of which led to a single amino acid change. Genetic diversity of a minicore collection of Chinese soybean landraces assessed by simple sequence repeat (SSR) markers and allelic variation at the GmTfl1 locus suggest that human selection for determinacy took place at early stages of landrace radiation. The GmTfl1 allele introduced into a determinate-type (tfl1/tfl1) Arabidopsis mutants fully restored the wild-type (TFL1/ TFL1) phenotype, but the Gmtfl1 allele in tfl1/tfl1 mutants did not result in apparent phenotypic change. These observations indicate that GmTfl1 complements the functions of TFL1 in Arabidopsis. However, the GmTfl1 homeolog, despite its more recent divergence from GmTfl1 than from Arabidopsis TFL1, appears to be sub-or neofunctionalized, as revealed by the differential expression of the two genes at multiple plant developmental stages and by allelic analysis at both loci.comparative genomics | domestication | diversification | point mutation

An electrode system is described for the near-simultaneous application and measurement of translational, levitational and rotational forces induced by AC electric fields, and this has been used to investigate the differences in the AC electrodynamics of viable and non-viable yeast cells. A new approach to the theoretical modelling of the experimental data has enabled these differences to be quantified in terms of changes in the conductivity of the cytoplasmic membrane and cell interior. The results are considered to have potentially important biomedical and biotechnological applications.

Brassica species, including crops such as cabbage, turnip and oilseed, display enormous phenotypic variation. Brassica genomes have all undergone a whole-genome triplication (WGT) event with unknown effects on phenotype diversification. We resequenced 199 Brassica rapa and 119 Brassica oleracea accessions representing various morphotypes and identified signals of selection at the mesohexaploid subgenome level. For cabbage morphotypes with their typical leaf-heading trait, we identified four subgenome loci that show signs of parallel selection among subgenomes within B. rapa, as well as four such loci within B. oleracea. Fifteen subgenome loci are under selection and are shared by these two species. We also detected strong subgenome parallel selection linked to the domestication of the tuberous morphotypes, turnip (B. rapa) and kohlrabi (B. oleracea). Overall, we demonstrated that the mesohexaploidization of the two Brassica genomes contributed to their diversification into heading and tuber-forming morphotypes through convergent subgenome parallel selection of paralogous genes.

Crustacea, the subphylum of Arthropoda which dominates the aquatic environment, is of major importance in ecology and fisheries. Here we report the genome sequence of the Pacific white shrimp Litopenaeus vannamei, covering ~1.66 Gb (scaffold N50 605.56 Kb) with 25,596 protein-coding genes and a high proportion of simple sequence repeats (>23.93%). The expansion of genes related to vision and locomotion is probably central to its benthic adaptation. Frequent molting of the shrimp may be explained by an intensified ecdysone signal pathway through gene expansion and positive selection. As an important aquaculture organism, L. vannamei has been subjected to high selection pressure during the past 30 years of breeding, and this has had a considerable impact on its genome. Decoding the L. vannamei genome not only provides an insight into the genetic underpinnings of specific biological processes, but also provides valuable information for enhancing crustacean aquaculture.

We describe here the use of the xCELLigence system for label-free and real-time monitoring of cell -viability. The xCELLigence system uses specially designed microtiter plates containing interdigitated gold microelectrodes to noninvasively monitor the viability of cultured cells using electrical impedance as the readout. The continuous monitoring of cell viability by the xCELLigence system makes it possible to distinguish between different perturbations of cell viability, such as senescence, cell toxicity (cell death), and reduced proliferation (cell cycle arrest). In addition, the time resolution of the xCELLigence system allows for the determination of optimal time points to perform standard cell viability assays as well as other end-point assays to understand the mode of action. We have used the WST-1 assay (end-point viability readout), the cell index determination (continuous monitoring of viability by xCELLigence), and the DNA fragmentation assay (end-point apoptosis assay) to systematically examine cytotoxic effects triggered by two cytotoxic compounds with different cell-killing kinetics. Good correlation was observed for viability readouts between WST-1 and cell index. The significance of time resolution by xCELLigence readout is exemplified by its ability to pinpoint the optimal time points for conducting end point viability and apoptosis assays.

A real-time cell electronic sensing (RT-CES) system was used for label-free, dynamic measurement of cell responses to cytotoxicants. Cells were grown onto the surfaces of microelectronic sensors, which are comprised of circle-on-line electrode arrays and are integrated into the bottom surfaces of the microtiter plate. Changes in cell status such as cell number, viability, morphology, and adherence were monitored and quantified by detecting sensor electrical impedance. For cell quantification and viability measurement, the data generated on the RT-CES system correlated well with those from the colorimetric (MTT) assay. For cytotoxicity assessment, cells growing on microelectronic sensors were treated with different cytotoxicants, such as arsenic, mercury, and sodium dichromate. The dynamic responses of the cells to the toxicants were continuously monitored by the RT-CES system. On the basis of the IC50 values, the RT-CES system displays an equal sensitivity to the neutral red uptake assay at specific time points. Furthermore, because the RT-CES system provides real-time information regarding the state of cell morphology and adhesion in addition to cell number, we were able to discern a previously unreported effect of arsenic on NIH 3T3 cells prior to cell death. Also, using the RT-CES system, we were able to monitor cytotoxicity effects that occur within a minute of compound addition. Taken together, the RT-CES system allows for real-time, continuous monitoring and quantitative recording of the whole assay process and provides new insight into the cell-toxicant interaction.

The separation and purification of human blood cell subpopulations is an essential step in many biomedical applications. New dielectrophoretic fractionation methods have great potential for cell discrimination and manipulation, both for microscale diagnostic applications and for much larger scale clinical problems. To discover whether human leukocyte subpopulations might be separable by such methods, the dielectric characteristics of the four main leukocyte subpopulations, namely, B- and T-lymphocytes, monocytes, and granulocytes, were measured by electrorotation over the frequency range 1 kHz to 120 MHz. The subpopulations were derived from human peripheral blood by magnetically activated cell sorting (MACS) and sheep erythrocyte rosetting methods, and the quality of cell fractions was checked by flow cytometry. Mean specific membrane capacitance values were calculated from the electrorotation data as 10.5 (+/- 3.1), 12.6 (+/- 3.5), 15.3 (+/- 4.3), and 11.0 (+/- 3.2) mF/m2 for T- and B-lymphocytes, monocytes, and granulocytes, respectively, according to a single-shell dielectric model. In agreement with earlier findings, these values correlated with the richness of the surface morphologies of the different cell types, as revealed by scanning electron microscopy (SEM). The data reveal that dielectrophoretic cell sorters should have the ability to discriminate between, and to separate, leukocyte subpopulations under appropriate conditions.