Genomics, proteomics, glycomics – modern biology is typified by the “omic” fields in which analyses of large quantities of genes, proteins or sugars are carried out, instead of looking at individual molecules. This can give the outward appearance that such high throughput studies always examine an organism’s entire set of molecules. However that is untrue, at least for proteins and sugars, because the appropriate tools to do so are not yet available.

The working group led by Ruedi Aebersold, ETH Zurich Professor at the Institute of Molecular Systems Biology (1) , in collaboration with researchers at the Institute for Systems Biology in Seattle, the University of California in Los Angeles and the Cellzome Company, have succeeded in improving methods for the analysis of proteins. The study, which will be published in the scientific journal “Nature Biotechnology” in February (2) , could pave the way for analyses which will then record an organism’s complete protein set.

The starting point for the scientists’ new paper was the discovery that, contrary to expectations, the different peptides, i.e. the fragments of the proteins being examined, occur with unequal frequency when analysed by mass spectrometry. This means that various peptides are more or less suitable for the identification of their parent protein. The researchers demonstrated this phenomenon again, based on their own examples. This also revealed that on average three different peptides are sufficient to identify 95 percent of proteins. This in turn implies that proteome analyses can be performed with a small number of proteotypic peptides. But what makes a peptide a protein identifier?

To answer this question, the scientists collected as many properties of the individual protein-identifying peptides as possible. For this they used more than 600,000 peptides which they generated on four proteomic platforms – i.e. different proteome analysis methods. This allowed 16,000 proteotypic peptides to be determined purely empirically from 4,030 yeast proteins. Finally a 1,000-dimensional properties vector including parameters such as charge or hydrophobicity was created for each peptide. By using pattern recognition methods it was then possible to reduce the properties space to the five dimensions having the greatest power of discrimination.

Complex apparatus and analyses are already a characteristic feature of proteome analysis. Thanks to ETH Zurich research, the analysis of all of an organism’s proteins at once is coming within reach.

Next the researchers examined their identification algorithm. This proved successful in determining diagnostically valuable peptides for known proteins, but was also usable when searching for new peptides corresponding to proteins not previously recorded. Astonishingly, this revealed that no linear relationship exists between protein length and the number of “diagnostic” peptides. The fact that the identification tool functions regardless of the different proteome platforms predicts that the new algorithm will be widely applicable

Ruedi Aebersold concludes that “Our findings enable a slimmer, faster method.” This is because previously the methods had been highly redundant and consequently very expensive. The identification of protein-specific peptides, which is now easier to carry out, will greatly improve protein quantification. Aebersold explains that it is now possible to label such peptides and to use known amounts of them as a reference.

The scientist intends to use the new algorithm in the future in model organisms to discover peptides from proteins that were never until now covered in proteome analyses. “The aim of this joint project by ETH Zurich and Zurich University within Systems X (3) is to determine for the first time the really complete proteome of various species including Drosophila melanogaster, Caenorhabditis elegans and Arabidopsis thaliana.”