High Throughput Screening (HTS), with all its strengths and limitations, is still the single-best way to discover novel interesting molecules in drug discovery. Thomas Kodadek of Scripps Florida has an interesting article on screening in the latest issue of Nat. Chem. Biol which is a special issue on chemical probes.
Kodadek talks about the very different properties required for drugs and probes and the limitations and unmet needs in current HTS strategies. He focuses on mainly two kinds of screening; functional assays and binding assays. The former can consist of phenotypic screening wherein one is only interested in a particular cellular response. This is more useful for drugs. However for probes, target selectivity is important and one must have knowledge of the target. HTS hits can hit all kinds of protein targets, thus making it hard to find out if your compounds are being selective. Mutagenesis and siRNA studies can shed light on target selectivity but this is not easy to do.
One of the possible solutions Kodadek suggests to circumvent the problem of gauging selectivity is to use binding assays instead of functional assays. He notes a pretty clever idea used in binding assays; that of throwing in cell extracts with miscellaneous proteins that could mop up greasy, non-selective compounds. This strategy cannot be easily used in functional assays. Binding assays are also typically less expensive than functional assays.
There is also a discussion of some of the very practical problems associated with screening. Screening typically has low hit rates and more importantly, hits from screening are not leads. You usually need a dedicated team of synthetic chemists to make systematic SAR modifications to these hits to optimize them further. As the author says, few synthetic chemists wish to serve as SAR facilities for their biologist colleagues. Plus it is not easy to lure industrial chemists to serve this function in academia (although the present economic climate may have made this easier). Thus, biologists with no synthetic background need to be able to make at least some modifications to their hits. For this purpose Kodadek suggests the use of modular molecules with easily available building blocks which can be cheaply and easily connected together by relatively inexperienced chemists; foremost in his recommendations are peptoids, N-substituted oligoglycines which are biologically active and easy to synthesize. Thus, if libraries for screening are enriched in such kinds of molecules, it could make it easy for biologists without access to sophisticated synthetic chemists and facilities to cobble together leads. Of course this would lead to a loss of diversity in the libraries, but that's the tradeoff necessary for going down the long road from hit to lead.
Lastly, Kodadek briefly talks about prospects for screening in academia. Academic drug discovery is gradually becoming more attractive with the recent long lull in industry. However academic scientists are typically not very familir with the post-synthesis optimization of drugs including optimization of metabolic properties, bioavailability and PK. Academic scientists who can pursue such studies or partner with DMPK contracting companies may be paid back their dues.
One topic which Kodadek does not mention is virtual screening (VS). VS can complement HTS and at least some studies indicate that the rate of success in VS can match, if not exceed, that in HTS. In addition, new ligand-based methods which use properties such as molecular shape to screen for compounds similar to given hits can also valuably complement HTS follow up studies.
Screening is still the best bet for discovering new drugs, but hit rates are typically still very low (1% would be a godsend). Only a concerted effort at designing libraries, ensuring selectivity and synthetic accessibility will make it easier.
Kodadek, T. (2010). Rethinking screening Nature Chemical Biology, 6 (3), 162-165 DOI: 10.1038/nchembio.303