Data-driven computational drug discovery has recently exploded; yet, compiling the necessary data poses serious difficulties. Biological systems are too complex for a single dataset to adequately describe, with genes, proteins, diseases, and drugs forming dense, indecipherable networks of interactions. Such complex and multi-faceted information can hardly be retrieved through a simple query on a public database. Formulas of chemical agents may be stored in one repository, more formulas in another, and information such as attributes of target proteins, molecular interactions, and drug side effects might be scattered in a number of databases. Drug researchers address this challenge in three main ways: First, they acquire data from multiple sources and repositories, and then integrate them through computational means. Second, they exploit text mining techniques, combing through biomedical literature for diverse information. Third, they combine mined text with database querying, locating whatever data they need for each research project. Here, we present a treatment on this topic, examining how scientists across academia and industry integrate multi-source data, what kind of data they use, what trends in approaches and best practices have been emerging, and how setbacks such as data imbalance are being tackled.