Bioprospecting is a catch-all term for activities including discovery, acquisition, and utilization of novel biomaterials. This has historically been a controversial activity, often leading to unregulated commercialization of fauna (e.g., plants and medicinals) from third world countries for the benefit of commercial interests [Pros/Cons of Bioprospecting]. However, as a term in Molecular Biology, it reflects the growing need to discover new types of protein and nucleic acid parts, which can be used in biotechnology and basic research. The advent of multiple Next-Generation Sequencing technologies since 2006 now provides depth of information into the entire genomes (Metagenomics) of species previously inaccessible to basic research. 
Although not planned, one of the great examples of Bioprospecting is the story of Green Fluorescent Protein (GFP), a protein that has had a profound impact on every major field in modern biology. Originally isolated and characterized by Osamu Shimomura in the 1960's and 1970's from jellyfish and sea pansies, it was a mere oddity that conferred the eery bioluminescence of certain deep sea creatures. However, the subsequent cloning of the gene by Martin Chalfie and improvement into enhanced GFP by Roger Tsien made it into one of the modern workhorses in biology. This 40 year journey earned Shimomura, Chalfie, and Tsien the 2008 Nobel Prize in Chemistry. [History of GFP]
Metagenomics uses Next Generation Sequencing Technologies (e.g., Whole Genome Shotgun Sequencing (WGS), Roche 454, Illumina, ABI Solid) to completely sample the genomes of mixed microbial communities, generating an unbiased view of genomic sequence space. Estimates have suggested that greater than 99% of all microbes are unculturable in the lab and inaccessable to traditional laboratory analysis. Thus, these Next Generation Sequencing approaches allow for analysis of microbes that are small percentages of a microbial community. The current explosion in various Metagenomic projects (340 current projects, 1990 samples [GOLD database]) permits for entirely in silico approaches to identifying new gene families, with potential as parts in Synthetic Biology.
Craig Venter and his Yacht
In the early 2000's, the J. Craig Venter Institute set as one of its goals to sequence the genomic diversity in the oceans. Craig Venter used his personal yacht, the Sorcerer II, to traverse the Earth's oceans, taking samples of oceanic life and sequencing using Whole Genome Shotgun Sequencing. From this adventure, they uncovered 6 million proteins (double the current database), which consisted of 1,700 clusters of gene families with no known homology. The data also revealed homology for 6,000 unknown ORF families (ORFan). They found that a very high proportion of new genes belonged to viruses (likely marine phage), which current databases had underrepresented. 
Recent Successes of Bioprospecting using Metagenomics (Targeted Metagenomics)
A useful approach to Bioprospecting new genes involves either functional screening or pure sequence screening in what is called Targeted Metagenomics. This involves either challenging microbiota to a particular activity, or looking for specific families of genes. Both types of Targeted Metagenomic screens have led to new antibiotic resistance genes, cold-adaptive rRNA's, and cellulosic enzymes, to name just a few. .
Typical Targeted Metagenomic Pipeline
Cellulosic Biomass degrading genes found in Cow Rumen
Plant polysaccharides such as cellulose are not broken down with enyzmes found in mammals, but species such as Ruminents (cows) carry symbiotic bacteria that perform this job. These microbes cannot be cultured in lab. However, acquisition of the enzymes used to break down cellulose could be used to generate biofuel from easily grown plants like grass. Here, Mattias Hess and colleagues used Metagenomic analysis to identify 51 enzymes active in breaking down polysaccharides. The authors isolated microbes from a nylon bag filled with switchgrass placed inside a fistula created into a cow's rumen. Various Next-Gen Sequencing technologies were used to generate 268 Giga-basepairs of sequences. From the various organisms, they predicted 27,755 putative polysaccharide enyzymes, of which 43% had less than 50% similarity to any known sequence. The authors expressed and tested 90 of these genes based on similarity to glycoside hyrdolase domains, which identified 51 active enzymes. The enyzmes were active against various substrates used as biofuel crops. Finally, the authors generated 15 draft genomes of new microbial species. 
Uses of New Parts
New genes found from Ruminant microbes are being used for generating fuel from biomass, proteins which have thus far been unknown to man. Genes identified in more extromophilic bacteria and archaea may be useful in metabolism of inorganic compounds. They may be useful in new genetic circuits in biotech applications. Finally, the genes will ultimately be useful as scaffolds for directed evolution studies, to generate new functions.
First, many of the current Next-Gen Sequencers are limited by their short-reads and need to perform emPCR, which can introduce bias. The coming introduction of single-molecule long read sequencers, such as that by Pacific Biosciences may alleviate some of these limitations. Finally, it is unlikely that genes found in nature will cover all the uses humanity may come up with. Since nature settles for genes that function "well-enough", this may be inadequate when humans are so concerned with efficiency and preoccupied with perfection.
- BiotechLettReview2010 pmid=20495950
- EnviroMicroReview2011 pmid=21366818
- PLOSbio2007 pmid=17355171
- Science2011 pmid=19181843