Publication Date

Spring 2011

Degree Type

Master's Project

Degree Name

Master of Science (MS)


Computer Science


Drug discovery depends on scientists finding similarity in molecular fingerprints to the drug target. A new way to improve the accuracy of molecular fingerprint folding is presented. The goal is to alleviate a growing challenge due to excessively long fingerprints. This improved method generates a new shorter fingerprint that is more accurate than the basic folded fingerprint. Information gathered during preprocessing is used to determine an optimal attribute order. The most commonly used blocks of bits can then be organized and used to generate a new improved fingerprint for more optimal folding. We thenapply the widely usedTanimoto similarity search algorithm to benchmark our results. We show an improvement in the final results using this method to generate an improved fingerprint when compared against other traditional folding methods.