Multiprotein complexes and metalloproteins largely determine the metabolic potential of an organism. The genes that encode multiprotein complexes and metalloproteins in any organism are largely unknown. Predicting what these genes are from genome sequence data is virtually impossible due to great diversity of amino acids sequences that are involved in protein oligomerization and metal binding. In this dissertation, an experimental method is presented that circumvents this problem by the direct purification of multiprotein complexes and metalloproteins from native microbial biomass. The biomass of a model microorganism, Pyrococcus furiosus, was fractionated by non-denaturing multistep chromatography and chromatographic fractions were analysed by mass spectrometry to identify proteins that co-eluted in common fractions. Potential multiprotein complex identification was based on the cofractionation of proteins encoded by adjacent genes in the P. furiosus genome. A total of 106 potential heteromeric multiprotein complexes were identified, of which only 20 were already known or predicted. One of them, PC-81 (PF1838/PF1837) was of interest because of its homology to two previously purified heterotetrameric (alpha2 beta 2) enzymes termed ACS1 (PF1540/PF1787) and ACS2 (PF1540/PF1837). Since PC-81 and ACS2 had a common -subunit (PF1837), we recombinantly expressed in E. coli all ten ACS complexes from the five subunits and two subunits encoded in the P. furiosus genome. All ten enzymes were active with varying but overlapping substrate specificities. Metalloproteins were identified in a similar manner except that fractions were analyzed by ICP-MS in addition to peptide mass spectrometry. Of the 343 metal peaks identified during biomass fractionation, 158 were not associated with a known metalloprotein. Two metalloproteins were purified to homogeneity. One of them, PF0086 or alanyl-tRNA synthetase COOH terminus had been predicted to contain zinc. However, it co-purified with nickel rather than a zinc peak through three chromatography steps. The other, PF1972 or anaerobic ribonucleotide reductase activase, is a known iron protein that very surprisingly co-purified with molybdenum as well as iron through five chromatography steps. This method therefore identified one new nickel and one new molybdenum-containing protein from P. furiosus, in addition to approximately 100 new multiprotein complexes, thereby demonstrating the general applicability of this approach.