An intentional approach to managing bias in general purpose embedding models.

Advances in machine learning for health care have brought concerns about bias from the research community; specifically, the introduction, perpetuation, or exacerbation of care disparities. Reinforcing these concerns is the finding that medical images often reveal signals about sensitive attributes in ways that are hard to pinpoint by both algorithms and people. This finding raises a question about how to best design general purpose pretrained embeddings (GPPEs, defined as embeddings meant to support a broad array of use cases) for building downstream models that are free from particular types of bias. The downstream model should be carefully evaluated for bias, and audited and improved as appropriate. However, in our view, well intentioned attempts to prevent the upstream components-GPPEs-from learning sensitive attributes can have unintended consequences on the downstream models. Despite producing a veneer of technical neutrality, the resultant end-to-end system might still be biased or poorly performing. We present reasons, by building on previously published data, to support the reasoning that GPPEs should ideally contain as much information as the original data contain, and highlight the perils of trying to remove sensitive attributes from a GPPE. We also emphasise that downstream prediction models trained for specific tasks and settings, whether developed using GPPEs or not, should be carefully designed and evaluated to avoid bias that makes models vulnerable to issues such as distributional shift. These evaluations should be done by a diverse team, including social scientists, on a diverse cohort representing the full breadth of the patient population for which the final model is intended.

Brain-wide representations of behavior spanning multiple timescales and states in C. elegans.

Changes in an animal's behavior and internal state are accompanied by widespread changes in activity across its brain. However, how neurons across the brain encode behavior and how this is impacted by state is poorly understood. We recorded brain-wide activity and the diverse motor programs of freely moving C. elegans and built probabilistic models that explain how each neuron encodes quantitative behavioral features. By determining the identities of the recorded neurons, we created an atlas of how the defined neuron classes in the C. elegans connectome encode behavior. Many neuron classes have conjunctive representations of multiple behaviors. Moreover, although many neurons encode current motor actions, others integrate recent actions. Changes in behavioral state are accompanied by widespread changes in how neurons encode behavior, and we identify these flexible nodes in the connectome. Our results provide a global map of how the cell types across an animal's brain encode its behavior.

Reaction-Based Scalable Inorganic Patterning on Rigid and Soft Substrates for Photovoltaic Roofs with Minimal Optical Loss and Sustainable Sunlight-Driven-Cleaning Windows.

Recently developed fabrication methods for inorganic patterns (such as laser printing and optical lithography) can avoid some patterning processes conducted by conventional etching and lithography (such as substrate etching and modulation) and are thereby useful for applications in which the substrates and materials must not be damaged during patterning. Simultaneously, it is also necessary to develop facile and economical methods producing inorganic patterns on various substrates without requiring a special apparatus while attaining the above-mentioned advantages. The present study proposes a reaction-based method for fabricating inorganic patterns by immersing substrates coated with a colloidal nanosheet into an aqueous solution containing inorganic precursors. Silica and TiO2 patterns spontaneously developed during the conversion of each inorganic precursor. These patterns were successful on rigid and flexible substrates. We fabricated these patterns on a wafer-sized silicon and large flexible poly(ethylene terephthalate) film, suggesting the scalability. We fabricated a biomimetic pattern on both sides of a glass window, as a photovoltaic roof, for minimal optical losses to maximally present photovoltaic effects of a solar cell. The TiO2 pattern on glass window exhibits sustainable sunlight-driven-cleaning activity for contaminants. The method could provide a platform for economical high-performance inorganic patterns for energy, environmental, electronics, and other areas.

Hiding the Rabbit: Using a genetic algorithm to investigate shape guidance in visual search.

During visual search, attention is guided by specific features, including shape. Our understanding of shape guidance is limited to specific attributes (closures and line terminations) that do not fully explain the richness of preattentive shape processing. We used a novel genetic algorithm method to explore shape space and to stimulate hypotheses about shape guidance. Initially, observers searched for targets among 12 random distractors defined, in radial frequency space, by the amplitude and phase of 10 radial frequencies. Reaction time (RT) was the measure of "fitness." To evolve toward an easier search task, distractors with faster RTs survived to the next generation, "mated," and produced offspring (new distractors for the next generation of search). To evolve a harder search, surviving distractors were those yielding longer RTs. Within eight generations of evolution, the method succeeds in producing visual searches either harder or easier than the starting search. In radial frequency space, easy distractors evolve amplitude × frequency spectra that are dissimilar to the target, whereas hard distractors evolve spectra that are more similar to the target. This method also works with naturally shaped targets (e.g., rabbit silhouettes). Interestingly, the most inefficient distractors featured a combination of a body and ear distractors that did not resemble the rabbit (visually or in spectrum). Adding extra ears to these distractors did not impact the search spectrally and instead made it easier to confirm a rabbit, once it was found. In general, these experiments show that shapes that are clearly distinct when attended are similar to each other preattentively.

BET inhibitors repress expression of interferon-stimulated genes and synergize with HDAC inhibitors in glioblastoma.

BACKGROUND: The development of rational combination therapies is key to overcome inherent treatment resistance of glioblastoma (GBM). We aim at identifying new druggable targets by disturbing GBM cells with inhibitors of bromodomain and extra-terminal motif (BET) proteins to reveal cancer-relevant vulnerabilities that may sensitize to a second drug. BET proteins are epigenetic modulators and have been associated with proto-oncogene overexpression in cancer. METHODS: A GBM-derived sphere-line was treated with the BET inhibitor (BETi) JQ1 over a time-course of 48 hours, followed by RNA-sequencing. Four chromatin marks were investigated by chromatin immunoprecipitation followed by sequencing (ChIP-seq). Signatures of interest were functionally validated in vitro and in orthotopic xenografts. Combination therapies were evaluated for synergistic effects. RESULTS: Cancer-relevant pathways significantly modulated by JQ1 comprised interferon alpha (IFN-α) response genes and response signatures to histone deacetylase inhibitors (HDACi). The IFN-signature was reminiscent of a GBM-derived IFN-signature comprising CD274 (PD-L1). Functional pathway analysis suggested that JQ1 was acting directly on the transcriptional level of IFN-response genes and not via the canonical JAK/STAT pathway. This was in line with JQ1 modulated expression and BRD4 and Pol II occupancy at IFN-signature genes, supporting a direct mechanistic interaction. Finally, we showed that combining HDACi with JQ1 acts synergistically in reducing cell viability of GS-lines. CONCLUSIONS: Our approach identified BETi-induced vulnerabilities in cancer-relevant pathways, potentially amenable to synergistic combinatorial therapy, such as combination with HDACi. The direct inhibitory effect of BETi on IFN-responsive genes in GBM cells, including CD274, indicates modulation of the tumor immune landscape and warrants further studies.

Myticusin-beta, antimicrobial peptide from the marine bivalve, Mytilus coruscus.

We isolated and purified an antimicrobial peptide (AMP) from the mantle of the hard-shelled mussel, Mytilus coruscus. The peptide was purified through C18 reversed-phase high-performance liquid chromatography, and displayed antibacterial activity. Total molecular mass of 11,182 Da was determined using matrix-assisted laser desorption ionization time-of-flight mass spectrophotometry. The N-terminal 23-amino acid sequence of its purified peak was obtained through Edman degradation, revealing 82% identity with myticusin-1 of M. coruscus. Complete sequence of the target peptide was determined through cDNA cloning and rapid amplification of cDNA ends. The complete sequence comprised 574 bp with a 387-bp open reading frame (ORF) encoding 24 amino acids of a signal peptide and 104 amino acids of a mature peptide, which was named myticusin-beta. Furthermore, we discovered two novel isoforms of myticusin-beta. We constructed and expressed recombinant myticusin-beta, which displayed antimicrobial activity against gram-positive (Bacillus cereus, Bacillus subtilis, Clostridium perfringens, Staphylococcus aureus, Streptococcus iniae, Streptococcus mutans) and gram-negative bacteria (Escherichia coli, Pseudomonas aeruginosa, Vibrio alginolyticus, Klebsiella pneumoniae). Purified recombinant myticusin-beta also showed anti-parasitic activity at various concentrations. A short AMP analog was designed and synthesized based on the sequence of myticusin-beta, with markedly improved antimicrobial activity. Expression of myticusin-beta was detected in the mantle at the highest level, followed by hemocytes. The results obtained in this work suggest that myticusin-beta is an immune-related AMP of M. coruscus and an effective alternative to antibiotics.

Intravenous immunoglobulin (IVIg) acts directly on conventional T cells to suppress T cell receptor signaling.

Intravenous immunoglobulin (IVIg) therapy is widely used to treat autoimmune and infectious disorders. Despite the clinical efficacy of IVIg therapy, its precise immunosuppressive mechanisms remain unclear. Here, we provide evidence that IVIg acts directly on T cells to suppress their activation upon T cell receptor (TCR) ligation. IVIg suppressed the proliferation of murine splenocytes upon stimulation with anti-CD3 antibody and T cell-tropic mitogens. These immunosuppressive effects of IVIg were still intact against purified T cells, and the depletion of naturally-occurring regulatory T cells (nTreg) had no effect on T cell regulatory activity. Instead, we found that IVIg negatively regulated TCR signaling; IVIg co-stimulation impaired IκB degradation, nuclear translocation of the nuclear factor of activated T cells (NFAT), and the activation of mitogen-activated protein kinase (MAPK, Erk1/2). These results suggest an additional new immunosuppressive role of IVIg, which acts directly on conventional T cells to suppress the TCR signaling pathway.

Effect of collector speed and flow rate on morphology of Er doped TiO2 nanofibers.

The 0.5 mol% Er3+ doped TiO2 (Er(3+)-TiO2) nanofibers were synthesized by a sol-gel derived electrospinning and subsequent calcination for 3 h at 500 degrees C in air. The calcined fibers were examined to evaluate the effect of collector speed and flow rate on morphology of the fibers. The dynamic viscosity and surface tension of precursor solution were 34 cP and 22.7 mN/m, respectively. The Er(3+)-TiO2 nanofibers were electrospun horizontally on the drum rotated at 100-500 rpm and flow rate of 0.2-0.5 mL/h under a DC voltage of 10 kV. The grounded collector is a stainless mandrel placed 12 cm away from the tip of the needle. Beads were observed for the nanofibers prepared at flow rates from 0.2 mL/h to 0.5 mL/h when the collector speed was 100 rpm. The nanofibers increased in diameter slightly from 150 nm to 190 nm as the flow rate was raised from 0.2 mLh to 0.5 mL/h. No beads were found at the collector speed of above 300 rpm when the flow rate was 0.2 mL/h. The optimized flow rate and collector speed of the nanofibers were determined to be in the range of 0.2-0.3 mL/h and 300-400 rpm, respectively. Uniform, smooth and continuous fibers with diameters of 150 to 170 nm were detected. Crystallite size determined by the Scherrer formula was about 6 nm. It can be concluded that the collector speed and the flow rate are influential on the morphology of the Er(3+)-TiO2 nanofibers. The Er(3+)-TiO2 nanofibers, prepared at 0.2 mL/h and 300 rpm, had typical absorption peaks located at 490, 523 and 654 nm, corresponding to the transitions from 4I15/2 to 4F7/2, 2H11/2 and 4F9/2, respectively. The Er(3+)-TiO2 nanofibers showed enhanced photoresponses under visible light.

Genetic differentiation among populations and color variants of sea cucumbers (Stichopus japonicus) from Korea and China.

The Far Eastern sea cucumber, Stichopus japonicus, is a favored food in Eastern Asia, including Korea, Japan, and China. Aquaculture production of this species has increased because of recent declines in natural stocks and government-operated stock release programs are ongoing. Therefore, the analyses of genetic structure in wild and hatchery populations are necessary to maintain the genetic diversity of this valuable marine resource. In addition, given that sea cucumber color affects market price, with the rare, possibly reproductively isolated, red type being the most valuable, an understanding of the genetic structure and diversity in color variation of green and red types is necessary. We analyzed the genetic structure of wild and hatchery-produced green type S. japonicus from Korea and China, and wild red type from Korea using 9 microsatellite makers. The number of alleles per locus ranged from 11 to 29 across all populations. The mean allele numbers of the green types from Korea (10.6) and China (10.1) were similar, but differed slightly from that of the red type (9.1). Pairwise multilocus F(ST) and genetic distance estimations showed no significant differences between the green types from Korea and China, whereas the differences between the green and red types were significant. This was clearly illustrated by a UPGMA dendrogram, in which the two close subclusters of green types were completely separated from the red type. In addition, the allele frequencies of the green and red types were significantly different. Assignment tests correctly assigned 100% (quality index 99.97%) of individuals to their original color types and demonstrated the feasibility of microsatellite analysis for discrimination between color types.

Effect of steric hindrance on carbon dioxide absorption into new amine solutions: thermodynamic and spectroscopic verification through solubility and NMR analysis.

Acid gas absorption technology is of great importance in these days for the prevention of global warming and the resulting worldwide climate change. More efficient process design and development for the removal of acid gases has become important, together with the development of new absorbents as one of urgent areas of research in addressing global-warming problems. In the present work, aqueous solutions of 2-amino-2-hydroxymethyl-1,3-propanediol (AHPD), a sterically hindered amine, has been examined as a potential CO2 absorbent and compared with the most commonly used absorbent, monoethanolamine (MEA) solution, through equilibrium solubility measurements and 13C NMR spectroscopic analyses. The solubilities of CO2 in aqueous 10 mass % AHPD solutions were higher than those in aqueous 10 mass % MEA solutions above 4 kPa at 298.15 K, but below 4 kPa, the solubility behavior appeared to be the opposite. The solubility difference between these two solutions increased with the CO2 partial pressures above the crossover pressure. Equilibrated CO2-MEA-H2O and CO2-AHPD-H2O solutions at various CO2 partal pressures ranging from 0.01 to 3000 kPa were analyzed by 13C NMR spectroscopy to provide a more microscopic understanding of the reaction mechanisms in the two solutions. In the CO2-amine-H2O solutions, amine reacted with CO2 to form mainly the protonated amine (AMH+), bicarbonate ion (HCO3-), and carbamate anion (AMCO2-), where the quantitative ratio of bicarbonate ion to carbamate anion strongly influenced the CO2 loading in the amine solutions. A profusion of bicarbonate ions, but a very small amount of carbamate anions, was identified in the CO2-AHPD-H2O solution, whereas a considerable amount of carbamate anions was formed in the CO2-MEA-H2O solution. AHPD contains more hydroxyl groups than nonhindered MEA, and hence, the chemical shifts in its 13C NMR spectra were strongly influenced by the solution pH values. In contrast, MEA appeared to be insensitive to pH. The strong interrelations among CO2 solubility, CO2 partial pressure, bulkiness of the amine structure, and pH identified through the present experimental investigations can provide basic guidelines for finding new potential organic absorbents, including specifically designed amine chemicals.