Most pro and eukaryote genomes encode hundreds of enzymes of unknown function; finding what they do is a critical task for postgenomic biology. Such enzymes cannot generally be components of core biosynthetic or degradative pathways because such functions would have been uncovered already by biochemical genetics. What then are they? Mounting evidence implicates many of them in ‘metabolite repair’, i.e. in reversing damage done to metabolites by unwanted enzymatic side-reactions or chemical degradation. Thus, since metabolites are under constant chemical attack and enzymes make wasteful and toxic catalytic errors, it follows that efficient functioning of metabolic networks demands a support system dedicated to metabolite repair. This system has been glimpsed by classical biochemistry, genetics, and metabolomics but most of it remains hidden. We therefore propose to dissect the metabolite repair system by combining chemical biology, comparative genomics, and metabolomics using bacterial models and plants.