"Given the huge numbers of possible compounds, and the relatively tiny fraction of those we can screen, why does high-throughput screening even work at all?" asks Derek Lowe at In the Pipeline. A recent paper from the Shoichet group at UCSF compared the contents of different screening libraries to what natural products generally look like, finding that the libraries appear to lean toward having compounds with structures similar to natural products, which, Lowe says, "we can reasonably assume have generally evolved to have some sort of biological activity."