After four meetings built around the tools and technologies used in proteomics, GeneProt co-founder Denis Hochstrasser and his fellow organizers of the Fifth Siena meeting decided the time had come to focus on proteomics applications.
Between two-hour lunches featuring Tuscan wine and cuisine, conference attendees did get a true flavor of the range of proteomics analyses, from applications in discovering biomarkers for heart transplant rejection to investigating the muscle proteins responsible for Tibetans’ ability to thrive at high altitude. In fact, the number of talks describing advancements in our understanding of biological function would overflow the cup of any knowledge-thirsty scientist.
But from a technology standpoint, several advancements and challenges emerged out of the din of clinking glasses: Researchers continue to make incremental advances to 2D gel electrophoresis and create novel chemical tags for simplifying protein mixtures; protein arrays still suffer from problems with specificity; and making sense of huge amounts of mass spectrometry data is just plain difficult.
A Common Platform for Comparing Protein IDs?
Beginning with the latter, Ruedi Aebersold of the Institute for Systems Biology spoke on two strategies under development in his laboratory for simplifying the process of converting data from mass spectrometers into protein identifications. “To make an impact on biology, we need to deliver high-throughput, high-quality data to biologists,“ he said. The problem is that proteomics researchers spend 75 percent of their time trying to validate their data, he added.
To address this problem, Aebersold described a two-step process for data validation currently in press in the journal Analytical Chemistry. As a means to automate mass spec data analysis and provide a common platform for comparing data obtained from different spectrometers, Aebersold’s group devised a method for applying multivariate statistics to validated protein identification data obtained from mass spectrometry experiments, and using it for newly collected data.
In the first step, Aebersold’s group created a score based on the accuracy of a predicted protein identification, and fit curves to the two distributions corresponding to the true and false predictions. Applying this model in the second step, Aebersold’s team used the scores from newly collected data to determine the probability that a given prediction is correct. “The end point is to compare data from different mass spectrometers, collected in different labs, using different database search algorithms,“ he said.
Aebersold also described a strategy for reducing the number of redundant peptide sequencing analyses. To avoid sequencing peptides already identified in previous runs, he described a technique for creating a library of synthetic peptides to use as markers in a protein identification experiment. By applying one form of the ICAT reagent to peptides obtained from a biological sample, and the other form of the reagent to the known synthetic peptides, a combined mass spectrometry analysis of the two samples would indicate which of the peptides derived from the biological sample were not contained in the library, or array, of synthetic peptides, he said. Because these unexpected peptides would appear in mass spectra as singlets, a researcher could pinpoint only those peptides for secondary sequencing analysis. For further details, he directed attendees to http://www.systemsbiology.org/research/software/proteomics.
Ahhh, The Challenge of Protein Arrays…
In his first talk as CEO of the Plasma Proteome Institute, a non-profit based in Washington, DC, former Large Scale Biology chief scientific officer Leigh Anderson spoke on the diagnostic potential of proteins coursing through the plasma proteome, and the lengths researchers must travel before that potential translates into clinical benefit. The plasma proteome is both the richest source of diagnostic markers and its most elusive, he said. Detecting a protein at the lower end of the concentration range, he added, is equivalent to finding “one bottle of wine in Lake Geneva.“
Furthermore, of the 300-odd plasma proteins currently validated for use as diagnostic markers, almost none were discovered using proteomics, he said. And given the current rate at which new FDA-approved diagnostic protein assays are being approved, in the future, that number will creep asymptotically toward zero. “Whatever we’re doing in proteomics is absolutely not transferring through to diagnostics,“ he said, “and that’s something we should all think about trying to address.“ [In a later talk, Scott Patterson, Celera’s vice president for proteomics, said the literature contains so few proteins identified by proteomics because they are all locked up in company databases.]
But it was not all doom and gloom. While Anderson certainly hopes to address this problem through his Plasma Proteome Institute, Dolores Cahill of the Max-Planck Institute of Molecular Genetics in Berlin also spoke on her progress developing arrays of denatured proteins derived from human fetal brain cDNA libraries and expressed in E. coli. As a first step, Cahill has employed oligonucleotide fingerprinting to eliminate the redundancies from an initial set of 37,000 human brain cDNAs, bringing the current total to over 11,000.
Cahill has arrayed these proteins onto both PVDF membranes and glass slides, and performed experiments to test their specificity using antibodies to GAPDH and heat shock proteins. The results, she indicated, leave something to be desired from antibody specificity. In one example, an antibody to heat shock proteins bound to 18 of its 24 theoretical epitopes, but also bound to six others non-specifically. During the Q&A session, Anderson and Cahill assailed the lack of antibody specificity, calling it “alarming,“ and “frightening,“ respectively, although Cahill noted that the antibodies seem to be epitope-specific, if not protein-specific.
Onward Separation Chemistry!
In one of a series of talks describing improvements to protein separation techniques, Bengt Bjellqvist of Amersham Biosciences presented his company’s putative solution to cysteinyl-related streaking in 2D gels. Using a proprietary reagent described as DeStreak, which oxidates cysteinyl groups to produce mixed disulfides, Bjellqvist showed gel images with significantly less visible streaks. Upon the conclusion of his talk, Amersham arranged for meeting organizer Hochstrasser to present Bjellqvist with an award for his accumulated contributions to gel electrophoresis technology.
After presenting data on his application of proteomics to chemoresistance in cancer cells, Pranav Sinha of the Charité University Hospital in Berlin described his collaborative work with ProteoSys, of Mainz, Germany, in developing ultra-large gel formats of up to one meter in length and 20 cm in width. Sinha is using the large gel formats with various pH gradients to separate human prostate tumor proteins, and said he expects to present his data at the upcoming ICES meeting in Glasgow, Scotland.
In a twist on traditional multidimensional liquid chromatography, Martin Larsen of the University of Southern Denmark described a novel approach for using micro columns packed with graphite powder to retain posttranslationally modified proteins for MDLC/MS analysis. Although most researchers would opt for 2D gels to analyze modified proteins, Larsen showed that his technique is a suitable alternative because the graphite powder micro columns are more hydrophobic than standard reverse-phase columns, thereby limiting losses of modified peptides.
Applied Biosystems’ Dale Patterson presented a demonstration of his company’s new cleavable ICAT reagents, which allow peptide pairs to co-elute from liquid chromatography separation and fragment more conveniently during mass spectrometry analysis, he said. In one experiment, Patterson’s results indicated that cleavable reagents allowed 83 percent coverage of b- and y- ions under MS analysis, compared with 54 percent by the standard ICAT reagents. Patterson also claimed that analyzing protein samples by both electrospray and MALDI ionization provides a significantly larger number of protein identifications. [Al Burlingame, of UCSF’s mass spectrometry laboratory, later upheld this view during a Q&A, while lamenting ABI’s “canned response,“ to the tittering of the audience.]