To develop more effective therapies to treat human diseases, a better method of finding the biological targets and modes of action of new compounds is needed. Target predictions have traditionally been made by comparing a new compound's molecular structure to that of known compounds. In many cases this method does not accurately predict a chemical's function since "small chemical changes in an active molecule can render it either nearly or completely inactive or increase its activity dramatically" (Eckert and Bajorath, 2007). Further, prediction by structural comparison has limited application; it can only be used on chemicals with established structures and only identifies new compounds that are structurally similar to known compounds.

A majority of existing drugs have been discovered by identifying the active ingredient of traditional medicines. More recent techniques of drug discovery screen a library of compounds for effectiveness in treating a single disease. However, this method requires re-screening the library when searching for treatments for other diseases; a critical barrier to expediting and scaling drug discovery. Screening efficiency is particularly important since advances in robotic chemical synthesis and the search for natural products from the oceans are rapidly increasing the size of drug candidate libraries.

In contrast to current approaches which screen compounds for treatments for single disease; my research focused on creating screening methods that deliver a library of chemical fingerprints which can be used to find potential drug candidates for a multiplicity of diseases.

My work produced three screening methods that generate fingerprints useful for predicting a compound's mode of action: cytological profiling, D- Map, and BioSpace. All of these showed positive results towards solving the screening bottleneck. Finally, combining these approaches to integrate these various fingerprints could increase prediction accuracy of screening methods.