High throughput screening (HTS) has long played a vital role in drug discovery within the pharmaceutical industry, and its use in academia is growing too. These large-scale assays are used to screen thousands of compounds in search of hits – those sought after compounds that are used as starting points in the development of novel therapeutics.
Yet hits aren’t always what they seem. Time and again, certain molecules produce positive assay results against several different protein targets, but show no efficacy in follow-up studies. These false positives, known as Pan-Assay Interference Compounds or PAINS, are not due to target binding, but instead result from interference with the assay detection technology or non-specific interaction with the target protein.
Improving the reliability of high-throughput screening with computational PAINS filters
PAINS really are a pain for drug developers, as significant amounts of time and resource can be wasted attempting to develop them into drugs. Indeed, a number of recent high-profile failures highlight that too often these compounds are unmasked much later in development, once patents have been acquired and after the molecule has progressed through animal studies. And with the growing use of HTS outside of Big Pharma, some have expressed concern that PAINS are becoming more of a problem than ever before.
To improve the reliability of hit identification, computational filters have been proposed as a means of flagging up PAINS in compound screening libraries. This concept has gained much support within the scientific community and a number of web-based tools have emerged to help researchers identify and filter such compounds in screening libraries.
As a result of increased concern over the risks of progressing PAINS in drug development programs, it is increasingly common for researchers to deprioritize hits flagged with PAINS alerts in virtual screening assays prior to experimental HTS studies. Similarly, lead compounds generated from experimental HTS campaigns that trigger PAINS alerts have also been deprioritized for further development. Some scientific journals have even begun to recommend that these filters be used to screen lead compounds reported in manuscripts prior to being considered for publication.
Are PAINS filter deprioritizing leads and limiting discovery efforts?
A growing number of voices are now cautioning against an over-reliance on PAINS filters to deprioritize leads during drug discovery, with many questioning whether the they are unnecessarily putting large areas of chemical space off-limits – even discarding drugs approved by the FDA and available on the market.
Adding further fuel to the debate is a recent paper that re-examines the original study from which the PAINS alerts were derived. This new investigation notes that the original authors based their assessment of compound promiscuity on just six protein-protein interaction inhibition assays using a single detection technology, leading the authors of this new work to conclude that the filters may have limited extrapolative power.
Of course, there is no question over the fact that compounds demonstrating pan-assay interference are a significant problem in drug discovery, and one that the community should be alert to. But while a broader awareness of PAINS may help us better recognize the structural features associated with these ‘frequent hitters’, these latest findings suggest that blindly using computational filters to restrict primary biological screens without performing orthogonal experimental assays, could severely limit, rather than assist, the discovery of new drug molecules.
To find out more about PAINS and these latest findings, read our article in full.