Higuchi Biosciences Center - University of Kansas

Fall 2006 Science Talks
November 30, 2006

Abstracts 1-15 | Abstracts 16-30 | Abstracts 31-45 | Abstracts 46-54

1. Peripheral Insulin may be Protective in Nondiabetics with Early Alzheimer Disease
Jeffrey M. Burns, Joseph E. Donnelly, Heather S. Anderson, Matthew S. Mayo, Luke Spencer-Gardner, George Thomas, Benjamin B. Cronk, David Klima, Daniel Hansen, William M. Brooks
Alzheimer and Memory Center, Landon Center on Aging, University of Kansas School of Medicine, Kansas City, KS
Bureau of Child Research, University of Kansas, Lawrence, KS

Background: Accumulating evidence suggests that insulin and insulin signaling may be involved in the pathophysiology of Alzheimer’s disease (AD). The relationship between insulin-mediated glucoregulation and brain structure has not been assessed in individuals with AD.
Methods: Nondemented (Clinical Dementia Rating (CDR) 0, n = 31) and early-stage AD (CDR 0.5 and 1, n = 31) participants aged 65 and over had brain MRI to determine whole brain and hippocampal volume while 3-hour intravenous glucose tolerance tests were used to determine glucose and insulin area-under-the-curve (AUC). Linear regression models were used to assess the relationship of insulin and glucose with brain volume, cognition, and dementia severity in individual diagnostic groups controlling for age.
Results: In early AD participants, insulin AUC was positively related to whole brain (β = 0.66, p <0.001) and hippocampal volume (β = 0.42, p<0.05). Glucose AUC was also related to whole brain and hippocampal volume. These relationships were independent of age, gender, body fat, cardiorespiratory fitness, physical activity, cholesterol, and triglycerides. Insulin AUC, but not glucose, was associated with global cognitive performance in early AD subjects (β = 0.40, p = 0.04), with better cognitive performance in those with increased serum insulin. Dementia severity (CDR box score) was associated with insulin AUC (Pearson’s r = -0.40, p = 0.03). Glucose and insulin AUCs were not related to brain volume or cognitive performance in nondemented individuals.
Conclusions: Increased serum insulin levels may be protective in nondiabetics in the earliest clinical stages of AD. The association between reduced peripheral insulin levels and more severe AD-related brain atrophy, cognitive dysfunction, and dementia severity suggests that reduced insulin signaling may play a role in the pathophysiology of AD.

-top-

2. Cloning and Characterization of Genes Involved in Ecdysone-Dependent Morphogenesis in Drosophila
Liang Zhang, Stefani Fontana, Ty Beaver, Kistie Patch, Robert Ward
Department of Molecular Biosciences, University of Kansas, Lawrence, KS

Our lab is interested in understanding the mechanisms of hormone dependent morphogenesis, and we have been using leg eversion during Drosophila metamorphosis as a model. Ecdysone is a steroid hormone that directs the major transitions during the Drosophila life cycle. broad (br) encodes a transcription factor that is induced by ecdysone and is necessary for leg eversion during metamorphosis. In a screen for dominant modifiers of a malformed leg phenotype associated with br1 (a weak loss of function mutation in broad), we identified mutations in at least 15 different genes. One of these mutations, Enhancer of broad 155 (E(br)155), shows complete embryonic lethality when homozygous, with poorly-differentiated cuticle and obvious defects in mid-embryonic morphogenesis. Specifically, 50% of the dead embryos have defects in dorsal closure, whereas 10% show defects in head involution. E(br)155 shows strong genetic interaction with br1 and Rho1. We initially mapped the enhancing mutation to 30A-F. Subsequent P-element based meiotic mapping of the lethal mutation indicated that the original mutation contained two closely mapped mutations, one in 26B and the other in 27D. After recombining the mutations apart we discovered that both mutations are required for the embryonic phenotypes and for the interactions with br1 and Rho1. We recently conducted an F2 mutagenesis screen and recovered 7 new alleles of the mutation in 26B and 3 new alleles of the mutation in 27D. We will present our characterization of these new alleles and our efforts in cloning the genes responsible for these mutations.

-top-

3. Differential Regional Relationship of Physical Fitness and Vascular Elasticity to Regional White Matter Volume in Normal Aging and Early Alzheimer’s Disease
George P. Thomas1,2, William Brooks2,3, Cary Savage3,5, Lee Graves4, Jeffrey M. Burns1,2
1Department of Neurology, University of Kansas Medical Center, Kansas City, KS
2Department of Integrative and Molecular Physiology, University of Kansas Medical Center, Kansas City, KS
3Hoglund Brain Imaging Center, University of Kansas Medical Center, Kansas City, KS
4Department of Endocrinology, University of Kansas Medical Center, Kansas City, KS
5Department of Psychiatry University of Kansas Medical Center, Kansas City, KS

Increased physical fitness levels appear to promote brain health in older adults, and may provide a degree of protection from cognitive decline. Changes in whole brain volume correlate with cognition in this same group. Vascular elasticity appears to be a logical candidate as a mediator of the fitness-brain health correlation. We examined the relationship between fitness, arterial elasticity and brain structure in a group of 55 demented and non-demented older adults. Our results show a strong correlation between overall white matter volume and both fitness and small artery elasticity. These correlations were not preset in gray matter.
To test the hypothesis that vascular elasticity, and ultimately fitness, is related to white matter changes in Alzheimer’s disease we performed a voxel-based analysis to look for differences in the regional distribution of the fitness-white matter volume relationship. We also looked at the arterial elasticity-white matter volume relationship using the same technique. Our results show that both fitness and arterial elasticity correlate with white matter volume in the frontal regions in non-demented subjects, while a different pattern was present in the demented subjects. In the latter, fitness correlated with white matter volume in the right parietal and cerebellar regions, while arterial elasticity correlated with white matter volume in the left parietal region. These distributions generally correspond to the patterns of brain atrophy seen in normal aging and in Alzheimer’s disease respectively. Our findings suggest that both fitness and arterial elasticity affect brain health in normal aging and early Alzheimer’s disease differently. The concept that the relationship between fitness and brain health is mediated by arterial elasticity is not directly supported however; in both the normal and demented subjects fitness appears to correlate with deep white matter, while arterial elasticity impacts peripheral white matter.

-top-

4. Synthesis of Oxazoline and Dihydrooxazine Libraries
Priyanka Chaudhry, Frank Schoenen, Jeffery Aubé
Department of Medicinal Chemistry, University of Kansas, Lawrence, KS
Chemical Methodologies and Library Development Center, University of Kansas, Lawrence, KS

The reactions of 1,2- and 1,3-hydroxyalkyl azides and aldehydes in the presence of Lewis acid result in the one-step construction of oxazolines and dihydroxazines respectively. The reaction was adapted to parallel synthesis by using a polymer-bound phosphine to scavenge the hydroxyalkyl azide, which was used in excess. Thus, a 60-member library of various disubstituted oxazolines and di- and tri-substituted dihydrooxazines has been generated.

Synthesis of Oxazoline and Dihydrooxazine Libraries

-top-

5. Systematic Electrokinetic Injection Investigation on Disposable Electrophoresis Microchips with Integrated Capacitively Coupled Contactless Conductivity Detection
Wendell Karlos Tomazelli Coltro, José Alberto Fracassi da Silva, Emanuel Carrilho
Instituto de Química de São Carlos, Universidade de São Paulo, São Carlos, SP, Brazil

In this report we present a miniaturized disposable device with integrated capacitively coupled contactless conductivity detection on polyester-toner (PT) electrophoresis microchips. The microchannels were fabricated by direct-printing process using a laser printer for toner deposition on polyester films. The microelectrodes were prepared on polyester films by sputtering deposition using toner layer deposited by laser-printer as masks for deposition. After the preparation of microchannels and microelectrodes, both polyester films were laminated together in order to provide the sealing of microfluidic device. The feasibility of the proposed microchip was evaluated by electrophoretic separation of potassium, sodium and lithium in microchannels of 150-µm wide and 6-µm deep. The electrodes were designed in an antiparallel configuration with 750-µm width and 750-µm gap between them. The best results were recorded in a frequency of 400 kHz and 10 Vpp. The analytes were successfully separated in less than 90 seconds. The detection limits found for K+, Na+ and Li+ ranged from 3 to 7 µmol L-1. A systematic electrokinetic injection investigation was carried out on these microchips. The addition of a clean-up step between consecutive injections improved the relative standard deviation of injection-to-injection from about 30% to less than 5%, increasing the reliability of the injections, which is highly desired for quantitative studies. The same chip has been used for more than 50 runs. The low cost of the proposed microsystem is one of the several advantages of this system. Besides the cost, the integrated microdevice eliminates the problem of manual alignment and gluing of the electrodes.

-top-

6. Temporal and Spacial Expression of Two Lysyl Oxidase-like Genes in Drosophila Melanogaster
1Matthew Culpepper, 1Robert E. Ward IV, 1Minae Mure, 2Brian Avery, 2Judith P. Klinman
1Department of Chemistry, University of Kansas, Lawrence, KS
2Department of Chemistry, University of California, Berkeley, CA

Lysyl oxidase (LOX) catalyzes the cross-linking of collagen and elastin and plays an essential role for extracellular matrix stabilization. Recently, three lysyl oxidase-like (loxl) genes have been identified in humans that contain scavenger receptor cysteine rich domains in addition to the highly conserved LOX catalytic domain. One of the loxl genes, hloxl2, has been implicated in invasion/metastasis of breast cancer. In order to understand the properties and functions of lysyl oxidase-like proteins, we have begun to investigate homologs in Drosophila. The Drosophila genome encodes two loxl genes (dloxl1 and dloxl2) and no LOX genes. Northern blot analyses revealed that dloxl1 is expressed at very high levels in the salivary glands of prepupae 4 to 8 hours after puparium formation, but is not expressed at detectable levels at any other stage of development. In contrast to dloxl1, dloxl2 is detectable only in adults, more so in male than female flies, and is expressed at substantially higher levels in the head than in the body. dloxl2 expression is significantly reduced in the heads of eya2 (eyes absent) mutant flies that completely lack compound eyes, suggesting that dloxl2 expression is confined to the eye. The distinct expression profiles of dloxl1 and dloxl2 suggests that they do not share a common function.

-top-

7. Creating a Mouse Model to Study the Function of Nuclear APC
Jamie Cunningham, Kristi Neufeld
Division of Biological Sciences, University of Kansas, Lawrence, KS

Mutations in the tumor suppressor adenomatous polyposis coli (APC) have been observed to be the initiating events in over 80% of all colorectal cancers. Inheriting a mutant allele of APC predisposes a person to colon polyps which eventually develop into malignant tumors. However, much remains unknown about the cellular function(s) of APC and how its loss of function results in tumorigenesis. APC can localize to numerous sites within the cell. APC binds to microtubule plus ends, is found at cell junctions, and has the ability to shuttle in and out of the nucleus. It is known that APC down regulates cytoplasmic β-catenin, but other functions, including the role of nuclear APC, remain unclear. We propose to study the function of nuclear APC in both embryo-derived stem (ES) cells and whole animals. To do this we will create a mouse model that lacks nuclear APC.
We have begun work creating ES cell lines expressing a full length APC that is unable to enter the nucleus. We currently have heterozygous cells and are making homozygous cell lines. This model system will be used to investigate the role of nuclear APC in regulating β-catenin activity, cellular proliferation and cellular differentiation. By injecting the mutant ES cells into mouse embryos we will develop animal models deficient in nuclear APC which can be used to investigate the role of nuclear APC in suppressing tumorigenesis under physiological conditions. While developing these models, we have also gathered baseline data from wild-type ES cells. Using confocal microscopy we have observed where APC, β-catenin and other proteins of interest are localized. These proteins seem to be found in similar locations in ES and in human colon tissue, supporting the validity of our mouse model as a tool for understanding how APC functions as a human tumor suppressor.

-top-

8. Synthesis of Cyclic Peptidomimetics Inspired by γ-Turns and Applications to Library Designs
Jeffrey Aubé, Erik Fenster
Chemical Methodologies and Library Development Center, University of Kansas, Lawrence KS

A series of peptidomimetics based on a γ-turn motif were prepared, in which N-protected piperidones were reacted with a selection of 2-hydroxyalkyl azides derived from common L-amino acids. Addition of a variety of nucleophiles to the initially formed iminium ethers afforded a series of diversified substituted 1,4-diazepinones. The applicability of this chemistry to the construction of 1,4-diazepinone libraries has also been investigated.

-top-

9. Development of a Separation-Based Sensor for Peroxynitrite
Celeste Frankenfeld
Department of Pharmaceutical Chemistry, University of Kansas, Lawrence, KS

Peroxynitrite (ONOO¯) is a quick-forming, fast-acting oxidative species that is generated from the reaction of nitric oxide (NO) and superoxide (O2¯˙). When cells are under stress, the concentration of ONOO¯ exceeds that of endogenous antioxidants, and ONOO¯ destroys DNA, cell membranes, and protein function through oxidation and nitration. ONOO¯ has been implicated in neurodegenerative, autoimmune, and heart diseases.
Under physiological conditions the half-life of ONOO¯ is less than a second and it is very difficult to detect directly. Therefore, most detection methods are indirect and focus upon the detection of ONOO¯ through oxidation or nitration of probes or trapping agents. However, microchip electrophoresis offers many advantages over other analytical techniques for the detection of ONOO¯. Most importantly, detection can occur directly, via amperometric electrodes that can be microfabricated and incorporated directly on-chip. Also advantageous is the electrophoretic separation that can be achieved within seconds, making it possible to separate and detect short-lived species, as well as monitor reactions with good temporal resolution. Common microchip substrates such as glass or PDMS are compatible for the growth of cells for bioanalysis, and can be used with a variety of solvent conditions including high pH.
In this presentation, a method for the separation and direct detection of ONOO- using microchip electrophoresis with amperometric detection will be reported. A number of different buffer systems, EOF modifiers, electrode materials, and pHs were evaluated for the separation of ONOO- from NO2¯ and H2O2. The final separation was accomplished using a run buffer consisting of 25 mM Na2HPO4 and 2 mM tetradecyltrimethylammonium bromide (TTAB). The optimized method was then used to follow the production of ONOO- from 3-morpholinosydnonimine (SIN-1). SIN-1 is the active metabolite of the heart drug molsidomine. Peaks were identified by changing the buffer pH (at pH 7 the halflife of ONOO- is less than a second and no peak is detected) and by determining the current ratio for the compounds at two different oxidation potentials.

-top-

10. Compound LibrariesBased on the Stemona Alkaloid Skeleton
Kevin J. Frankowski,a, b, Jeffrey Aubé,a, b
a Chemical Methodologies and Library Development Center, University of Kansas, Lawrence, KS
bDepartment of Medicinal Chemistry, University of Kansas, Lawrence, KS

The impressive antitussive activity of two Stemona alkaloids, neostenine and neotuberosemonine, inspired the synthesis of diversely-functionalized libraries based on the Stemona alkaloid core. Combining the Diels-Alder and azido Schmidt reactions provided access to the complex tricyclic skeleton of this class of alkaloids in an efficient and rapid manner. The tricyclic scaffolds were derivatized in parallel by either reductive amination or the Fischer indole synthesis. These compound libraries represent intriguing variations on the naturally-occurring Stemona alkaloids. Screening these libraries for specific opioid activity will begin to explore the intricate relationship between molecular architecture and biological activity of this class of alkaloids.

-top-

11. Improving Dynamic Range on the LTQ-FT Instrument Using Hhypothesis-Driven Mass Spectrometry
Nadezhda A. Galeva1, Yakov M. Koen2, Viktor S. Sharov3, Christian Schöneich3, Robert P. Hanzlik2, Todd D. Williams1
1Mass Spectrometry Laboratory, University of Kansas, Lawrence, KS
2Department of Medicinal Chemistry, University of Kansas, Lawrence, KS
3Department of Pharmaceutical Chemistry, University of Kansas, Lawrence, KS

Introduction Dynamic range is a challenge for mass spectrometric (MS) analysis of protein posttranslational modifications (PTM) and metabolites especially in trap instruments with gain control. PTM/metabolite revealing peptides are present in such small quantities that their MS signals are masked by abundant “native” peptides. Hypothesis-driven multiple-stage mass spectrometry (HMS-MS) [1] offers a strategy for experiments targeted to sensitive detection and identification of modified peptides and overcoming dynamic range limitations. We present a hypothesis-driven approach for studying nitration of tyrosine (Tyr) residues in Sarco/endoplasmic reticulum Ca-ATPase (SERCA) and bromobenzene metabolites in Glutathione S-transferase (GST) at cysteine (Cys) residues with LTQ-FT mass spectrometer. A discovery experiment for protein mapping is followed by an experiment with a look-up table populated by the hypothesis-modified peptides.
Method Samples of digested with trypsin proteins were introduced to a LTQ-FT mass spectrometer by capillary LC. The first data-dependent experiment was performed using dynamic exclusion with survey MS scans acquired in the FT-ICR and MS/MS scans acquired in the ion trap (IT). Peptides were identified by matching data against sequences of the target protein(s) using Sequest. For SERCA nitration, a look-up table was calculated by adding 45 amu to Tyr-containing peptides. The detection experiment using the look-up table was with survey MS scans acquired in the IT followed by narrow mass range MS scans using FT-ICR and MS/MS scans in the IT. Bromobenzene metabolites were localized to GST by radioactivity and were assumed to be on the best nucleophile, Cys.
Preliminary data Our hypotheses in the PTM studies were that all Tyr residues of SERCA were potential sites for nitration and for every modified tryptic peptide a corresponding unmodified peptide should be present in the digest. First, we performed a typical protein mapping experiment and used a Sequest search to identify all Tyr-containing SERCA peptides, including modifications such as methionine (Met) and Cys oxidation. We then calculated m/z values of all nitrated Tyr peptides for which unmodified precursors had been identified. Third, we used that mass list for precursor selection in a survey MS acquired in the linear IT scan over a mass range determined by the m/z values and using a low signal threshold to trigger MS/MS. The selected peptide ions were subjected to narrow mass range FT ICR and MS/MS simultaneously to allow unambiguous identification based on the exact mass and fragmentation pattern. The strategy for GST metabolites was essentially similar to the PTM project, with the first step to map the protein and identify unmodified tryptic peptides. The hypothesis for GST was that the metabolite would alkylate the most active nucleophile, Cys. The detection experiment was to acquire FT ICR over a narrow mass range and MS/MS spectra on selected ions with an assumption that delta mass between a metabolite and unmodified peptide would be less than a 300 amu. Therefore ion selection was operated not from a specific mass list but from a specific mass range. For example, for triply charged peptides containing Cys111 a survey IT-MS scan was performed over m/z 1050-1300 range. Specific fragmentation signatures such as enhanced fragmentation at the N-terminus of proline (Pro) residue allowed us to identify the family of peptides of interest. The level of modification was determined to be 0.9% (0.009 mol/mol).
Novel Aspect A new experimental design using hypothesis-driven mass spectrometry is proposed for identification of low abundant protein PTM and metabolites.
1.Anal.Chem.,2004, 76(15),4472-83

-top-

12. Processing ESI-MS1 Data Obtained on a Q-TOF Mass Spectrometer to Improve the Results of Protein AnalysisNadezhda A. Galeva1, Lei Jiang2, Asma Zaidi2, Mary L. Michaelis2, and Todd D. Williams1
1Mass Spectrometry Laboratory, University of Kansas, Lawrence, KS
2Department of Pharmacology and Toxicology, University of Kansas, Lawrence, KS

Introduction Protein mapping and identification in biological samples is routinely done using mass spectrometry (MS) and tandem mass spectrometry (MS/MS). Currently available commercial software for electrospray ionization (ESI) instruments, such as Q-TOF, employs MS/MS data for database searches/sequence analysis of proteolytic fragments, while the high quality MS1 data that are usually acquired during survey scans in a data-dependent experiment are not utilized for protein identification. The sensitivity and mass accuracy inherent in the MS1 data can be critical to obtaining useful information from samples of low protein abundance with poor quality MS/MS data. Here we present a way of processing and presenting ESI-MS1 data to complement and improve the results of database searches, protein mapping and quantification SILAC-labeled peptides.
Method   2D gel bands were tryptically digested. Peptides were separated on an RP column with a CapLC. Peaks were eluted into a Q-TOF-2™ mass spectrometer and analyzed in data-dependent fashion with dynamic exclusion. Precursor ions were selected in a survey scan with a low threshold and fragmented. ESI-MS/MS data were processed with ProteinLynx Global Server (PLGS) using the AutoMod workflow and search against SWISS-PROT database. LC ESI-MS1 data were processed by averaging time slices and charge deconvoluting using MaxEnt3. A peptide mass/intensity list was created for each raw file by combining the most intensive peaks from the MaxEnt spectra. The obtained peptide masses were used in a peptide mass fingerprint database search. Protein identifications were manually validated.
Preliminary Data   The first sample type we confronted was ESI-MS(MS) data for low-abundant protein samples from which we obtained few or poor quality ESI-MS/MS spectra. Peptide masses obtained by processing the ESI-MS1 data were used for fingerprinting searches against SWISS_PROT database. Usually there were enough masses for a search engine to return hits with high scores. Top ranked sequences for each sample were manually examined for presence of MS-tags derived from MS/MS spectra which allowed us to verify the identifications. Thus, information contained in the ESI-MS data was crucial for identification of low-abundant proteins and the MS/MS data was used to confirm peptide identities. The second sample type was 13C-labelled arginine (Arg-13C6) peptides in Stable Isotope Labeling with Amino acids in Cell culture (SILAC) experiments for quantitative proteomics. The ESI-MS1 data were used to find additional peptides of unlabeled-labeled peptide pairs from an already identified protein, and perform quantitative calculations using peptide intensities. Third, we used the ESI-MS data for more complete mapping of proteins of known sequence. Analysis of 173 residue protein resulted in an increase from 60% to 73% of protein coverage by processing both ESI-MS/MS and ESI-MS1data.
Novel Aspect   Using ESI-MS data rescued protein identification for low-abundant samples, enhanced coverage of known proteins, improved quantitative proteomics of SILAC peptides.

-top-

13. Expression of Collagen-like (GPP)n Pepitides in E. coli
Hidehiko Hirakawa, Julian Limburg
Department of Chemistry, University of Kansas, Lawrence, KS

Gly-Pro-Pro repetitive peptides are major components of procollagen and are hydroxylated by the collagen prolyl 4-hydroxylases. Gly-Pro-Pro repetitive peptides are model substrates for kinetic studies on P4Hs. However, commercially available Gly-Pro-Pro repetitive peptide is restricted to (Gly-Pro-Pro)10 and expensive due to solid phase synthesis. Peptides can be also obtained using E. coli expression system, but it is difficult to express short peptides. We show that a Gly-Pro-Pro repetitive peptide was expressed in the form of fusion protein with thioredoxin and the purified peptide was obtained by specific cleavage using a protease. A library of genes encoding repetitive Gly-Pro-Pro was prepared with synthetic oligonucleotides GGTCCACCGGGTCCACCG and CGGTGGACCCGGTGGACC using overlap elongation PCR method. From the two-stage screening (SDS-PAGE analysis and sequencing), we obtained 11 clones; 5, 6, 7, 8, 10, 13, 15, 16, 17, 19, and 23 repeats of Gly-Pro-Pro. A fusion protein of (Gly-Pro-Pro)19 peptide with thioredoxin was successfully expressed in E. coli. This protein is soluble and was easily purified with affinity chromatography and gel-filtration chromatography. The purified protein was cleaved with enterokinase at 37 °C overnight. The purified (Gly-Pro-Pro)19 peptide was obtained from gel-filtration chromatography. This peptide successfully worked as a substrate of human prolyl 4- hydroxylase.

-top-

14. Characterization and Structural Studies on D52 Tumor Protein
Yasufumi Yamamoto, Robert P. Hanzlik, Weijun Huang
Structural Biology Center, University of Kansas, Lawrence, KS

The D52 Tumor Proteins (TPD52) are small proteins of ~23kDa that are highly expressed in human breast cancer and other types of cancers, including ovarian, prostate and lung cancer. They exist in humans in three isoforms, TPD52, TPD52L1 and TPD52L2. Current evidence suggests that they form complexes with key molecules such as the phospholipid-binding protein annexin VI, MAL2, a raft-associated integral membrane protein of the MAL family, apoptosis signal-regulating kinase 1 (ASK1), as well as SNARE proteins syntaxin 1 and VAMP2. Evidence also suggests that the TPD52 proteins are involved in vital cell functions, such as vesicle trafficking, exocytotic secretion, calcium mediated signal transduction and cell proliferation. We have obtained the cDNAs for the D52 tumor proteins. We intend to make overexpression constructs for GST and intein fusion proteins, and purify the tumor proteins to homogeneity. We will then crystallize these proteins and determine their structures with X-ray crystallography.

-top-

15. Using Peptide Properties to Improve the Accuracy of Peptide Identification Through MS/MS and Database Search
Jianwen Fang1,2, Yinghua Dong1, Todd D Williams3, Gerald H Lushington1,4
1Bioinformatics Core Facility, University of Kansas, Lawrence, KS
2Information and Telecommunication Technology Center, University of Kansas, Lawrence, KS
3Mass Spectrometry Laboratory, University of Kansas, Lawrence, KS
4Molecular Graphics and Modeling Laboratory, University of Kansas, Lawrence, KS

Tandem mass spectrometry (MS/MS) with database search has emerged as the method of choice for the identification of proteins in high-throughput proteomics studies. However, currently this approach requires time-consuming manual validation in order to reduce false positives caused by the imperfect predictions of protein database search engines. Software tools that attempt to automatically validate the prediction have been developed. In this study we sought to improve these tools by investigating the potential use of peptide properties in the predictive models. Feature selection algorithms based on Random Forest and Support Vector Machine were used to identify peptide properties that may improve the predictive models. The improved model based on an optimized set of features reduced the number of false positives by 58% relative to the model used search engine scores alone. The sequence-dependent fragmentation and ionization patterns identified in the study can be incorporated into peptide sequencing and database search algorithms and improve current database search programs.

-top-

Higuchi Biosciences Center
University of Kansas
2099 Constant Avenue
Lawrence, KS 66047-2535
785-864-5183
hbc@ku.edu