California Institute of Technology

About: California Institute of Technology is a education organization based out in Pasadena, California, United States. It is known for research contribution in the topics: Galaxy & Redshift. The organization has 57649 authors who have published 146691 publications receiving 8620287 citations. The organization is also known as: Caltech & Cal Tech.

Inducible Foxp3+ regulatory T-cell development by a commensal bacterium of the intestinal microbiota

Abstract: To maintain intestinal health, the immune system must faithfully respond to antigens from pathogenic microbes while limiting reactions to self-molecules. The gastrointestinal tract represents a unique challenge to the immune system, as it is permanently colonized by a diverse amalgam of bacterial phylotypes producing multitudes of foreign microbial products. Evidence from human and animal studies indicates that inflammatory bowel disease results from uncontrolled inflammation to the intestinal microbiota. However, molecular mechanisms that actively promote mucosal tolerance to the microbiota remain unknown. We report herein that a prominent human commensal, Bacteroides fragilis, directs the development of Foxp3+ regulatory T cells (Tregs) with a unique “inducible” genetic signature. Monocolonization of germ-free animals with B. fragilis increases the suppressive capacity of Tregs and induces anti-inflammatory cytokine production exclusively from Foxp3+ T cells in the gut. We show that the immunomodulatory molecule, polysaccharide A (PSA), of B. fragilis mediates the conversion of CD4+ T cells into Foxp3+ Treg cells that produce IL-10 during commensal colonization. Functional Foxp3+ Treg cells are also produced by PSA during intestinal inflammation, and Toll-like receptor 2 signaling is required for both Treg induction and IL-10 expression. Most significantly, we show that PSA is not only able to prevent, but also cure experimental colitis in animals. Our results therefore demonstrate that B. fragilis co-opts the Treg lineage differentiation pathway in the gut to actively induce mucosal tolerance.

Gaussian Process Optimization in the Bandit Setting: No Regret and Experimental Design

Abstract: Many applications require optimizing an unknown, noisy function that is expensive to evaluate. We formalize this task as a multi-armed bandit problem, where the payoff function is either sampled from a Gaussian process (GP) or has low RKHS norm. We resolve the important open problem of deriving regret bounds for this setting, which imply novel convergence rates for GP optimization. We analyze GP-UCB, an intuitive upper-confidence based algorithm, and bound its cumulative regret in terms of maximal information gain, establishing a novel connection between GP optimization and experimental design. Moreover, by bounding the latter in terms of operator spectra, we obtain explicit sublinear regret bounds for many commonly used covariance functions. In some important cases, our bounds have surprisingly weak dependence on the dimensionality. In our experiments on real sensor data, GP-UCB compares favorably with other heuristical GP optimization approaches.

Mass and environment as drivers of galaxy evolution in SDSS and zCOSMOS and the origin of the Schechter function

Abstract: We explore the simple inter-relationships between mass, star formation rate, and environment in the SDSS, zCOSMOS, and other deep surveys. We take a purely empirical approach in identifying those features of galaxy evolution that are demanded by the data and then explore the analytic consequences of these. We show that the differential effects of mass and environment are completely separable to z ~ 1, leading to the idea of two distinct processes of "mass quenching" and "environment quenching." The effect of environment quenching, at fixed over-density, evidently does not change with epoch to z ~ 1 in zCOSMOS, suggesting that the environment quenching occurs as large-scale structure develops in the universe, probably through the cessation of star formation in 30%-70% of satellite galaxies. In contrast, mass quenching appears to be a more dynamic process, governed by a quenching rate. We show that the observed constancy of the Schechter M* and α_s for star-forming galaxies demands that the quenching of galaxies around and above M* must follow a rate that is statistically proportional to their star formation rates (or closely mimic such a dependence). We then postulate that this simple mass-quenching law in fact holds over a much broader range of stellar mass (2 dex) and cosmic time. We show that the combination of these two quenching processes, plus some additional quenching due to merging naturally produces (1) a quasi-static single Schechter mass function for star-forming galaxies with an exponential cutoff at a value M* that is set uniquely by the constant of proportionality between the star formation and mass quenching rates and (2) a double Schechter function for passive galaxies with two components. The dominant component (at high masses) is produced by mass quenching and has exactly the same M* as the star-forming galaxies but a faint end slope that differs by Δα_s ~ 1. The other component is produced by environment effects and has the same M* and α_s as the star-forming galaxies but an amplitude that is strongly dependent on environment. Subsequent merging of quenched galaxies will modify these predictions somewhat in the denser environments, mildly increasing M* and making α_s slightly more negative. All of these detailed quantitative inter-relationships between the Schechter parameters of the star-forming and passive galaxies, across a broad range of environments, are indeed seen to high accuracy in the SDSS, lending strong support to our simple empirically based model. We find that the amount of post-quenching "dry merging" that could have occurred is quite constrained. Our model gives a prediction for the mass function of the population of transitory objects that are in the process of being quenched. Our simple empirical laws for the cessation of star formation in galaxies also naturally produce the "anti-hierarchical" run of mean age with mass for passive galaxies, as well as the qualitative variation of formation timescale indicated by the relative α-element abundances.