Dragan Radulovic

Informatics Platform for Global Proteomic Profiling and Biomarker Discovery Using Liquid Chromatography-Tandem Mass Spectrometry

We have developed an integrated suite of algorithms, statistical methods, and computer applications to support large-scale LC-MS-based gel-free shotgun profiling of complex protein mixtures using basic experimental procedures. The programs automatically detect and quantify large numbers of peptide peaks in feature-rich ion mass chromatograms, compensate for spurious fluctuations in peptide signal intensities and retention times, and reliably match related peaks across many different datasets. Application of this toolkit markedly facilitates pattern recognition and biomarker discovery in global comparative proteomic studies, simplifying mechanistic investigation of physiological responses and the detection of proteomic signatures of disease.

Genetic basis of proteome variation in yeast

Proper regulation of protein levels is essential for health, and abnormal levels of proteins are hallmarks of many diseases. A number of studies have recently shown that messenger RNA levels vary among individuals of a species and that genetic linkage analysis can be used to identify quantitative trait loci that influence these levels. By contrast, little is known about the genetic basis of variation in protein levels in genetically diverse populations, in large part because techniques for large-scale measurements of protein abundance lag far behind those for measuring transcript abundance. Here we describe a label-free, mass spectrometry–based approach to measuring protein levels in total unfractionated cellular proteins, and we apply this approach to elucidate the genetic basis of variation in protein abundance in a cross between two diverse strains of yeast. Loci that influenced protein abundance differed from those that influenced transcript levels, emphasizing the importance of direct analysis of the proteome.

Genetic variation shapes protein networks mainly through non-transcriptional mechanisms.

Variation in the levels of co-regulated proteins that function within networks in an outbred yeast population is not driven by variation in the corresponding transcripts.

MglA Regulates Francisella tularensis subsp. novicida (Francisella novicida) Response to Starvation and Oxidative Stress

MglA is a transcriptional regulator of genes that contribute to the virulence of Francisella tularensis, a highly infectious pathogen and the causative agent of tularemia. This study used a label-free shotgun proteomics method to determine the F. tularensis subsp. novicida (F. novicida) proteins that are regulated by MglA. The differences in relative protein amounts between wild-type F. novicida and the mglA mutant were derived directly from the average peptide precursor ion intensity values measured with the mass spectrometer by using a suite of mathematical algorithms. Among the proteins whose relative amounts changed in an F. novicida mglA mutant were homologs of oxidative and general stress response proteins. The F. novicida mglA mutant exhibited decreased survival during stationary-phase growth and increased susceptibility to killing by superoxide generated by the redox-cycling agent paraquat. The F. novicida mglA mutant also showed increased survival upon exposure to hydrogen peroxide, likely due to increased amounts of the catalase KatG. Our results suggested that MglA coordinates the stress response of F. tularensis and is likely essential for bacterial survival in harsh environments.

Comparison of a Label-Free Quantitative Proteomic Method Based on Peptide Ion Current Area to the Isotope Coded Affinity Tag Method

Recently, several research groups have published methods for the determination of proteomic expression profiling by mass spectrometry without the use of exogenously added stable isotopes or stable isotope dilution theory. These so-called label-free, methods have the advantage of allowing data on each sample to be acquired independently from all other samples to which they can later be compared in silico for the purpose of measuring changes in protein expression between various biological states. We developed label free software based on direct measurement of peptide ion current area (PICA) and compared it to two other methods, a simpler label free method known as spectral counting and the isotope coded affinity tag (ICAT) method. Data analysis by these methods of a standard mixture containing proteins of known, but varying, concentrations showed that they performed similarly with a mean squared error of 0.09. Additionally, complex bacterial protein mixtures spiked with known concentrations of standard proteins were analyzed using the PICA label-free method. These results indicated that the PICA method detected all levels of standard spiked proteins at the 90% confidence level in this complex biological sample. This finding confirms that label-free methods, based on direct measurement of the area under a single ion current trace, performed as well as the standard ICAT method. Given the fact that the label-free methods provide ease in experimental design well beyond pair-wise comparison, label-free methods such as our PICA method are well suited for proteomic expression profiling of large numbers of samples as is needed in clinical analysis.

Proteomic Classification of Acute Leukemias by Alignment-Based Quantitation of LC–MS/MS Data Sets

Despite immense interest in the proteome as a source of biomarkers in cancer, mass spectrometry has yet to yield a clinically useful protein biomarker for tumor classification. To explore the potential of a particular class of mass spectrometry-based quantitation approaches, label-free alignment of liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) data sets, for the identification of biomarkers for acute leukemias, we asked whether a label-free alignment algorithm could distinguish known classes of leukemias on the basis of their proteomes. This approach to quantitation involves (1) computational alignment of MS1 peptide peaks across large numbers of samples; (2) measurement of the relative abundance of peptides across samples by integrating the area under the curve of the MS1 peaks; and (3) assignment of peptide IDs to those quantified peptide peaks on the basis of the corresponding MS2 spectra. We extracted proteins from blasts derived from four patients with acute myeloid leukemia (AML, acute leukemia of myeloid lineage) and five patients with acute lymphoid leukemia (ALL, acute leukemia of lymphoid lineage). Mobilized CD34+ cells purified from peripheral blood of six healthy donors and mononuclear cells (MNC) from the peripheral blood of two healthy donors were used as healthy controls. Proteins were analyzed by LC-MS/MS and quantified with a label-free alignment-based algorithm developed in our laboratory. Unsupervised hierarchical clustering of blinded samples separated the samples according to their known biological characteristics, with each sample group forming a discrete cluster. The four proteins best able to distinguish CD34+, AML, and ALL were all either known biomarkers or proteins whose biological functions are consistent with their ability to distinguish these classes. We conclude that alignment-based label-free quantitation of LC-MS/MS data sets can, at least in some cases, robustly distinguish known classes of leukemias, thus opening the possibility that large scale studies using such algorithms can lead to the identification of clinically useful biomarkers.

Weak Convergence of Smoothed Empirical Processes: Beyond Donsker Classes

Let X 1,…, X n be a sequence of independent random variables with common distribution P on the real line. We study smoothed empirical processes.\(\sqrt n \left( {{P_n} * {K_h} - P} \right){\left( f \right)_{f \in F}}\) based on a kernel estimate \({P_n} * {K_h}\), and indexed by a class of uniformly bounded functions F. Assuming that P has a smooth density, we demonstrate that the smoothed empirical process converges weakly to a tight Gaussian limit for classes F that are not necessarily P-Donsker. In addition, we provide some theory on the bootstrap estimate of the distribution of the process. In particular, we show that the bias inherent to the kernel density estimation method can consistently be captured by the bootstrap.

On the Bootstrap and Empirical Processes for Dependent Sequences

The goal of this paper is to describe recent advances in empirical processes theory and several bootstraps for dependent data. We will address both theory and applications. The first three sections deal with the heuristics, motivation, statement of results and applications. The last section is devoted to mathematical techniques behind the theory. Although some results presented here are new (bootstrap for Markov chains), this is not a research paper, and the presented proofs do not contain all the details.

Necessary and sufficient conditions for weak convergence of smoothed empirical processes

Let X1,...,Xn be a sequence of i.i.d. random variables with common distribution P on the real line. Assuming that P has a smooth density, we construct a histogram based estimator Pn,H and establish weak convergence of the empirical process under sharp conditions. If is a class of indicators of sets, then the conditions imposed are necessary and sufficient.

Renewal type bootstrap for Markov chains

In this paper we treat a renewal type of bootstrap for atomic Markov chains under minimal moment conditions on renewal times, i.e.Er2<∞. Three main results are: a) if a Markov chain satisfies the CLT for the mean then it also satisfies a bootstrap CLT; b) if a Markov chain satisfies a uniform CLT over classes of functions then, it also satisfies bootstrap uniform CLT with minimal condition on envelope functionF; c) we establish second order correctness for this procedure. All results are for “in probability” bootstrap and constitute the final word in this setting.