Abstract In the current data-intensive era, the traditional hands-on method of conducting scientific research by exploring related publications to generate a testable hypothesis is well on its way of becoming obsolete within just a year or two. Analyzing the literature and data to automatically generate a hypothesis might become the de facto approach to inform the core research efforts of those trying to master the exponentially rapid expansion of publications and datasets. Here, viewpoints are provided and discussed to help the understanding of challenges of data-driven discovery. The Panama Canal, the 77-kilometer waterway connecting the Atlantic and Pacific oceans, has played a crucial role in international trade for more than a century. However, digging the Panama Canal was an exceedingly challenging process. A French effort in the late 19th century was abandoned because of equipment issues and a significant loss of labor due to tropical diseases transmitted by mosquitoes. The United States officially took control of the project in 1902. The United States replaced the unusable French equipment with new construction equipment that was designed for a much larger and faster scale of work. Colonel William C. Gorgas was appointed as the chief sanitation officer and charged with eliminating mosquito-spread illnesses. After overcoming these and additional trials and tribulations, the Canal successfully opened on August 15, 1914. The triumphant completion of the Panama Canal demonstrates that using the right tools and eliminating significant threats are critical steps in any project. More than 100 years later, a paradigm shift is occurring, as we move into a data-centered era. Today, data are extremely rich but overwhelming, and extracting information out of data requires not only the right tools and methods but also awareness of major threats. In this data-intensive era, the traditional method of exploring the related publications and available datasets from previous experiments to arrive at a testable hypothesis is becoming obsolete. Consider the fact that a new article is published every 30 seconds (Jinha, 2010). In fact, for the common disease of diabetes, there have been roughly 500,000 articles published to date; even if a scientist reads 20 papers per day, he will need 68 years to wade through all the material. The standard method simply cannot sufficiently deal with the large volume of documents or the exponential growth of datasets. A major threat is that the canon of domain knowledge cannot be consumed and held in human memory. Without efficient methods to process information and without a way to eliminate the fundamental threat of limited memory and time to handle the data deluge, we may find ourselves facing failure as the French did on the Isthmus of Panama more than a century ago. Scouring the literature and data to generate a hypothesis might become the de facto approach to inform the core research efforts of those trying to master the exponentially rapid expansion of publications and datasets (Evans & Foster, 2011). In reality, most scholars have never been able to keep completely up-to-date with publications and datasets considering the unending increase in quantity and diversity of research within their own areas of focus, let alone in related conceptual areas in which knowledge may be segregated by syntactically impenetrable keyword barriers or an entirely different research corpus. Research communities in many disciplines are finally recognizing that with advances in information technology there needs to be new ways to extract entities from increasingly data-intensive publications and to integrate and analyze large-scale datasets. This provides a compelling opportunity to improve the process of knowledge discovery from the literature and datasets through use of knowledge graphs and an associated framework that integrates scholars, domain knowledge, datasets, workflows, and machines on a scale previously beyond our reach (Ding et al., 2013).