No abstract

Residual variance and the signal-to-noise ratio are important quantities in many statistical models and model fitting procedures. They play an important role in regression diagnostics, in determining the performance limits in estimation and prediction problems, and in shrinkage parameter selection in many popular regularized regression methods for high-dimensional data analysis. We propose new estimators for the residual variance, the ℓ 2 -signal strength, and the signal-to-noise ratio that are consistent and asymptotically normal in high-dimensional linear models with Gaussian predictors and errors, where the number of predictors d is proportional to the number of observations n. Existing results on residual variance estimation in high-dimensional linear models depend on sparsity in the underlying signal. Our results require no sparsity assumptions and imply that the residual variance may be consistently estimated even when d > n and the underlying signal itself is non-estimable. Basic numerical work suggests that some of the distributional assumptions made for our theoretical results may be relaxed.

Previously, we reported that the time course for the rapid phosphorylation rate of mu-opioid receptor expressed in human embryonic kidney (HEK)293 cells did not correlate with the slow receptor desensitization rate induced by [D-Ala(2),N-MePhe(4), Gly-ol(5)]-enkephalin (DAMGO). However, others have suggested that receptor phosphorylation is the trigger for mu-opioid receptor desensitization. In this study, we demonstrated the relatively slow rate of receptor desensitization could be attributed partially to the recycling of internalized receptor as determined by fluorescence-activated cell-sorting analysis. However, the blockade of the endocytic and Golgi transport events in HEK293 cells with monensin and brefeldin A did not increase the initial rate of receptor desensitization. But the desensitization rate was increased by reduction of the mu-opioid receptor level with beta-furnaltrexamine (betaFNA). The reduction of the receptor level with 1 microM betaFNA significantly increased the rate of etorphine-induced receptor desensitization. By blocking the ability of receptor to internalize with 0.4 M sucrose, a significant degree of receptor being rapidly desensitized was observed in HEK293 cells pretreated with betaFNA. Hence, mu-opioid receptor is being resensitized during chronic agonist treatment. The significance of resensitization of the internalized receptor in affecting receptor desensitization was demonstrated further with human neuroblastoma SHSY5Y cells that expressed a low level of mu-opioid receptor. Although DAMGO could not induce a rapid desensitization in these cells, in the presence of monensin and brefeldin A, DAMGO desensitized the mu-opioid receptor's ability to regulate adenylyl cyclase with a t(1/2) = 9.9 +/- 2.1 min and a maximal desensitized level at 70 +/- 4.7%. Furthermore, blockade of receptor internalization with 0.4 M sucrose enhanced the DAMGO-induced receptor desensitization, and the inclusion of monensin prevented the resensitization of the mu-opioid receptor after chronic agonist treatment in SHSY5Y cells. Thus, the ability of the mu-opioid receptor to resensitize and to recycle, and the relative efficiency of the receptor to regulate adenylyl cyclase activity, contributed to the observed slow rate of mu-opioid receptor desensitization in HEK293 cells.

Penalized least squares procedures which directly penalize the number of variables in a regression model (L 0 penalized least squares procedures) enjoy nice theoretical properties and are intuitively appealing. On the other hand, L 0 penalized least squares methods also have significant drawbacks. For instance, implementing these procedures is NP-hard and computationally unfeasible when the number of variables is even moderately large. One of the challenges in implementing L 0 penalized least squares procedures is discontinuity of the L 0 penalty. We propose the seamless-L 0 (SELO) penalty, a smooth function on [0, ∞) which very closely resembles the L 0 penalty. The SELO penalized least squares procedure is shown to consistently select the correct model and is asymptotically normal, provided the number of variables grows slower than the number of observations. SELO is efficiently implemented using a coordinate descent algorithm. Tuning parameter selection is crucial to the performance of the SELO procedure. We propose a BIC-like tuning parameter selection method for SELO and show that it consistently identifies the correct model, while allowing the number of variables to diverge. Simulation results show that the SELO procedure with BIC tuning parameter selection performs very well in a variety of settings -outperforming other popular penalized least squares procedures by a substantial margin. Using SELO, we analyze a publicly available HIV drug resistance and mutation dataset and obtain interpretable results.

One of the fundamental goals of proteomics methods for the biological sciences is to identify and quantify all proteins present in a sample. LC-MS/MS-based proteomics methodologies offer a promising approach to this problem (1-3). These methodologies allow for the acquisition of a vast amount of information about the proteins present in a sample. However, extracting reliable protein abundance information from LC-MS/MS data remains challenging. In this work, we were primarily concerned with the analysis of data acquired using bottom-up label-free LC-MS/MS-based proteomics techniques where "bottom-up" refers to the fact that proteins are enzymatically digested into peptides prior to query by the LC-MS/MS instrument platform (4), and "label-free" indicates that analyses are performed without the aid of stable isotope labels. One challenge inherent in the bottom-up approach to proteomics is that information directly available from the LC-MS/MS data is at the peptide level. When a protein-level analysis is desired, as is often the case with discovery-driven LC-MS research, peptide-level information must be rolled up into protein-level information.Spectral counting (5-10) is a straightforward and widely used example of peptide-protein roll-up for LC-MS/MS data. Information experimentally acquired in single stage (MS) and tandem (MS/MS) spectra may lead to the assignment of MS/MS spectra to peptide sequences in a database-driven or database-free manner using various peptide identification software platforms (SEQUEST (11) and Mascot (12), for instance); the identified peptide sequences correspond, in turn, to proteins. In principle, the number of tandem spectra matched to peptides corresponding to a certain protein, the spectral count (SC), 1 is positively associated with the abundance of a protein (5). In spectral counting techniques, raw or normalized SCs are used as a surrogate for protein abundance. Spectral counting methods have been moderately successful in quantifying protein abundance and identifying significant proteins in various settings. However, SC-based methods do not make full use of information available from peaks in the LC-MS domain, and this surely leads to loss of efficiency.Peaks in the LC-MS domain corresponding to peptide ion species are highly sensitive to differences in protein abundance (13,14). Identifying LC-MS peaks that correspond to detected peptides and measuring quantitative attributes of these peaks (such as height, area, or volume) offers a promFrom the