The discovery of new drugs is a very important area of study in medicinal chemistry. Developing a drug is not an easy task, as much time and money are needed to undertake all steps required for the development and test of new drugs. Amid this context, chemoinformatics is the area that has the role of interfacing between chemistry and computing, assisting in the process of identifying potential new drugs, through machine learning techniques for classification. This article will present the difficulties of classification found in chemoinformatics and approach machine learning techniques that, applied in the context of chemoinformatics, assist in treating issues related to uncertainty in data labeling and unbalanced classes, as they are common problems when using data sets of a chemical nature.