Drug Screening of GPCR Using Active Learning

In drug screening, active compounds (actives) bound to targets are found in large collections ofcompounds. Typically, between several hundred thousand to a million compounds are examined.It is highly expensive to test all compounds in random (Random), therefore limited biochemical testsare performed in searching for actives. Consequently, chemists start with some initial tested set anditeratively choose sets of compounds that are closest to previously known actives, where the similarityis usually calculated using the Tanimoto coecient (Tanimoto). To obtain a high number of actives,we applied active learning to drug screening. One of the active learning approaches is \query bybagging (QBag)" proposed by Abe and Mamitsuka [1].