Using data from Astra Zeneca patents we looked at the correlation of calculated physical properties with IC50 values obtained after serial dilution with tip-based methods or via acoustic dispensing (using a Labcyte, Inc device). In addition we analyzed more complex methods for modeling the data using pharmacophores (Discovery Studio, Accelrys). While the paper goes into far more detail, the bottom line is that compounds were more active when using the acoustic dispensing approach and that the pharmacophore for this data more closely resembled the X-ray structure derived pharmacophores, and showed the importance of hydrophobic features.
While this is really only the beginning of the story, it is important to consider alongside a whole array of other issues that may impact your data. In addition making sure we annotate our data with as much information – such as how we dispensed the compounds, may be useful to others in the future that need to model or analyze the data.