High-throughput screening (HTS) is a common technique for both drug discovery and basic research, but researchers often struggle with how best to derive hits from HTS data. While a wide range of hit identification techniques exist, little information is available about their sensitivity and specificity, especially in comparison to each other. To address this, we have developed the open-source NoiseMaker software tool for generation of realistically noisy virtual screens. By applying potential hit identification methods to NoiseMaker-simulated data and determining how many of the pre-defined true hits are recovered (as well as how many known non-hits are misidentified as hits), one can draw conclusions about the likely performance of these techniques on real data containing unknown true hits. Such simulations apply to a range of screens, such as those using small molecules, siRNAs, shRNAs, miRNA mimics or inhibitors, or gene over-expression; we demonstrate this utility by using it to explain apparently conflicting reports about the performance of the B score hit identification method.
Availability and implementation: NoiseMaker is written in C#, an ECMA and ISO standard language with compilers for multiple operating systems. Source code, a Windows installer and complete unit tests are available at http://sourceforge.net/projects/noisemaker. Full documentation and support are provided via an extensive help file and tool-tips, and the developers welcome user suggestions.