Proteins are beastly to work with: they denature at the drop of a hat and have an annoying tendency to regulate their social interactions by post-translational modifications. Small wonder, then, that researchers wanting a high-throughput readout of cellular behaviour use DNA microarrays to look at messenger RNA levels instead, even though they don't necessarily correlate with protein activity. All that might be about to change: in the 8 September issue of Science, Gavin MacBeath and Stuart Schreiber report that they can make microarrays of functionally active proteins, and can use them to measure interactions with other proteins and small molecules.

Two hurdles had to be leapt: keeping the proteins active and getting them in the right orientation. A third goal was to make the technology compatible with existing microarray analysis tools. With hindsight, the solutions to these problems turned out to be laughably simple: use the gear that prints commercially available DNA microarrays, put 40% glycerol in your buffers to prevent dehydration of the nanolitre volumes applied, and coat your slides with a reagent that reacts with primary amines. This captures proteins by their amino termini or by surface-exposed lysine residues, so each protein gets stuck to the glass in a range of different orientations, one of which is almost bound to be the right way up. The slides are then quenched with bovine serum albumin (BSA) which, as well as blocking any unreacted groups, lowers background noise when the slides are probed with other proteins. These simple tricks have allowed MacBeath and Schreiber to print proteins at densities of 1,600 spots per square centimetre.

Now pick your favourite protein function. Do you want to find new protein–protein interactions? Or hunt for new substrates for your pet protein kinase? Or are you more interested in finding out what proteins your library of drug candidates binds to? The researchers did proof-of-principle experiments to show that all of these applications are feasible by flooding the slides with fluorophore-tagged proteins, kinase substrates in the presence of radiolabelled ATP, or synthetic ligands coupled to fluorescently labelled BSA. Although most of these experiments were done using a small number of arrayed protein spots, they also work in the context of a chip containing over 10,000 spots: a single spot of the FKBP12–rapamycin binding protein (FRB) can easily be located in a sea of protein-G spots when probed with a mixture of two fluorophore-tagged proteins — one binding to FRB, the other binding to protein G.

The greatest barrier to commercial availability of these protein microarrays will be purification of the proteins to put on them. Let's hope that the current trend in automation obviates the need for arrays of protein biochemists, cursing in cold rooms over jammed fraction collectors.