Proteins in vivo often function in complexes, and indeed, that?s how many individual structural biology efforts approach them. Not structural genomics efforts, though: For all their high-throughput methods, structural genomics pipelines typically treat proteins individually, in isolation. A paper released May 11 in PNAS could help bridge this gap.

The new method, developed by David Eisenberg of the University of California, Los Angeles, and colleagues combines a series of bioinformatics analyses that scour a genome for clues suggesting that two genes are functionally linked (that is, that they both participate in a particular complex, pathway, or biological function), with experimental methods to validate those predictions.

The group focused on two large, but poorly characterized protein families in the Mycobacterium tuberculosis genome, called PE and PPE. Of 17 PE and 11 PPE proteins tested in their study, only one was soluble in E. coli on its own, and that one was unfolded. But when they jointly expressed one member of each family on a polycistronic message in bacteria, the proteins formed a highly expressed, heterodimeric complex. The resulting crystal structure reveals why neither protein was soluble individually: both proteins expose an "extended apolar interface" that is only buried (and thus, stabilized) when the two proteins form a complex.

This is precisely the reason why approaching proteins piecemeal is so hit or miss: many proteins simply don?t behave well in isolation. That observation dovetails nicely with a lesson I learned a few months ago at Keystone -- that structural genomics approaches, especially of eukaryotic proteins, have high failure rates. Given the experience of Eisenberg et al. with the PE and PPE families, I wonder if perhaps this new approach might not go a long way toward improving the odds.