Plastic pipette tips are ubiquitous in today’s laboratories. Yet they are not without problems. Hydrophobic compounds sometimes stick to them, and some solvents can extract compounds from the plastic that could potentially influence enzymatic reactions.
a new study in PLoS ONE suggests that tip-based dispensers may be leading drug developers astray, both in terms of the hits that emerge from their chemical screens and the compounds they choose not to pursue (1).
“It gives me nightmares,” said study author Sean Ekins, who consults on computational modeling, drug discovery, and toxicology for Collaborations in Chemistry. He believes that the study calls into question much of the available data in chemical databases. “I bet 95 to 99 percent [of existing data] were done with pipettes. And if people use those data for building models or drug discovery, that can lead you in the wrong direction.”
The new study directly compares tip-based liquid dispensing with acoustic dispensing, a non-contact-based method that uses sound energy to eject nanoliter volumes into an inverted vessel positioned above the sample. The team used data from two AstraZeneca patents describing inhibitors of the receptor tyrosine kinase EphB4. The patents included both structural and IC50 data—which measure how effective a molecule is at inhibiting a function—for EphB4 inhibitors, collected using both tip-based and acoustic dispensing methods.
In all cases, the IC50 values determined using tip-based dispensing were higher, in some cases much higher (up to 276.5 times) than the values determined using acoustic dispensing, indicating that the inhibitors were effectively less potent. Furthermore, the ranking of those compounds also differed, meaning the most potent compounds identified by tip-based dispensing were not the same as by acoustic dispensing, and vice versa.As a result, analyses using tip-based dispensing identified completely different inhibitors than those using acoustic dispensing.
The team then went on to compute pharmacophores for EphB4 inhibition from these data. They found that the two dispensing methods produced entirely different outcomes: Tip-based dispensing identified only hydrogen-bonding elements, but acoustic dispensing also identified those with hydrophobic features, which actually agrees with published structural data of EphB4 complexed with inhibitor molecules.
The findings are “deeply disturbing,” writes Derek Lowe, an organic chemist and drug developer who blogs on drug discovery at In the Pipeline, in a May 3 post. Lowe suggests that the effect Ekins documents likely stems from what he calls the sticky-compound effect, in which some compounds such as taxol prefer the plastic of pipette tips and microtiter plates to the aqueous solution of an assay well.
“An entire drug discovery program could be derailed as medicinal chemists build their structure-activity relationship framework around red herring compounds, while new molecules with great potential are discarded at the high throughput screening stage,” said Andy Holt from the University of Alberta, a pharmacologist who has documented issues with plastic labware in peer-reviewed papers.
How big a problem this is for the pharmaceutical industry is unknown, in part because pharma companies don’t normally publish the screening and structural data that would be required to better measure the effect.
In addition to Ekins, the other stud authors were Antony Williams from the Royal Society of Chemistry and Joe Olechno from Labcyte, a company that develops acoustic droplet liquid handling systems.
In the end, the authors suggest that acoustic dispensing methods would improve high-throughput screening results. But Holt notes that “at present, costs may render such technology inaccessible to most academic laboratories.”
Another improvement would be inclusion of meta-data indicating the form of liquid dispensing in public chemical databases “to create larger data sets for eventually comparing different computational models in future.”
For now, Ekins said, “As a modeler I would be very, very cautious using tip-based data.”
1. S. Ekins, J. Olechno, and A.J. Williams. Dispensing processes impact apparent biological activity as determined by computational and statistical analyses. PLoS ONE, 8:e62325, 2013. (doi:10.1371/journal.pone.0062325)