The challenge of industrializing proteomics

To the editor:

Two special issues of Nature Biotechnology¹ and Nature² set out to highlight progress and identify the challenges that lie ahead for industrializing proteomic technologies. In both issues, a rosy picture was painted of the potential of mass spectrometric approaches (e.g., isotope-coded affinity tags; ICAT) to impact efforts to quantitatively investigate molecular physiology. Unfortunately, some of the real obstacles that currently prevent proteomic techniques (including ICAT) from delivering on its promise were understated, particularly as it applies to use in drug discovery. It is important to emphasize this view to manage expectations of such technology, considering that pharma is a top-end consumer and adopter of proteomics.

Quantitative proteomics has much to offer in understanding the physiological consequences played out at the molecular level following inhibition of a drug target. However, the challenges associated with applying quantitative proteomics in this setting are complex. Experiments are multivariate, involving tissues, drugs, time points and physiological models. A typical experiment may investigate two tissues, with two drugs, collecting three time points for five animals (that is, 60 samples). Such an experiment generates mountains of data, which once collected, must be prioritized, filtered, analyzed and assembled. In parallel, business drivers push projects forward, meaning that timelines for delivering results from proteomics (and other technologies) are painfully short and strict.

Overcoming these challenges are real concerns for the industrialization of proteomics. Unfortunately, in their current incarnation, quantitative mass spectrometric approaches are far too slow to be considered practical, efficient solutions for operating in such an environment. The problem lies not with the approach, rather, in transitioning the data into knowledge. As pointed out by Patterson³, “data analysis is the Achilles heel of proteomics” and “advances will require intensive efforts on the computational front.”

These are solvable problems that need attention to move this approach into the mainstream. In the meantime, for all their drawbacks, two-dimensional gels remain a reliable, although undesirable option, to address throughput for quantitative proteomics. Identifying proteins and measuring their expression are only the initial steps along a far lengthier path in using the outputs from proteomics for decision-making in drug discovery. Ascribing functional attributes to proteins of interest is key, and assembling the reams of data produced in a proteomics experiment (and complementary parallel experiments) into pathways of physiology is essential to generate testable hypotheses. In combating these challenges we are only just testing the waters. Our knowledge of protein function is frustratingly weak, and it is only through tenacious, laborious efforts that we can expect to expand it. Solutions to these challenges need to be at the forefront of efforts to industrialize proteomics.