Advancing biomedical data science literacy

Biomedical data science is a rapidly growing interdisciplinary field. The combination of large-scale statistics, medical sciences, machine learning, and informatics has the potential to completely transform health. It is also one that greatly benefits from a diversity of ideas and is enriched by welcoming students from a wide range of backgrounds from statistics and computer science to molecular biology and anthropology. It is imperative, therefore, to be able to offer a core training plan that fluidly accommodates this diversity. The development of a rigorous training program that satisfies these needs is a great passion of mine. I am deeply invested in training individuals whose technical acumen is matched by a keen scientific curiosity. Researchers equipped with the ability to identify the important research questions as well as the skills to answer them will be the next generation of scientific leaders.

Over the last 10 years, I have led two courses and guest lectured in dozens of others. For five years I was the course director for our introductory graduate course, Introduction to Computer Applications in Health Care. This course was a series of guest lectures that gave the students a broad overview of the field of informatics from clinical and cognitive informatics to bioinformatics and genome sequencing.

The second course was one I developed from the ground up, Translational Bioinformatics, and is a hybrid of didactic lectures, flipped-classrooms, and peer-instruction. Learning materials center around a practical and hands-on approach to biomedical informatics applications in the translational sciences (e.g. population health, pharmaceutical sciences, molecular disease etiology, and human genetics). The focus is on how to become a good interdisciplinary scientist. To take ideas from computer science and statistics and map them to problems in molecular biology and medicine.

I have continually experimented with the structure, content, and format of this course. In its latest iteration it begins with a journal club where the students select from a curated list of recently published top translational bioinformatics papers. The journal club has two important effects. First, it highlights the importance for self-guided critical analysis. Second, it emphasizes the central role that communication and presentation will play in the course and thus in their scientific careers.

The second phase of the course is a series of lectures on translational bioinformatics topics – genetics and genomics, biomedical machine learning, biostatistics, chemical informatics, pharmacogenomics, drug repurposing, to name a few. During this time the students submitted research topics of interests and methods with which they are familiar. I curate these submissions into a matrix of challenge questions, data resources, and data science techniques.

In the third and final phase of the course, we mix and match methods and research topics together and discuss how they fit, or do not fit. For example, we may take a method developed for gene expression analysis and attempt to apply it to electronic health records. We then discuss the assumptions of the method and how the new data may violate those assumptions. We discuss how we could adapt the method to more appropriately fit the problem at hand or how the data might be processed to better fit the method. This final phase of the course is extremely rewarding. First, it leads the discussion into directions that are completely unexpected, and it gives students with expertise in a particular time to shine during class. Since the topics are so wide ranging, I cannot expect to be an expert on all of them. Therefore, whenever possible I ask the students to teach us about their submitted topics. This year, the students have led discussions on network analysis methods, feature engineering using deep learning, multiple sclerosis, Alzheimer’s disease, and breast cancer, to name a few.

Lastly, throughout the course the students are completing writing assignments as homework. The overall student product for the course is a 2,500-word translational bioinformatics research proposal on a drug repurposing project of their choice. Each homework builds toward this goal with a series of writing exercises and drafts culminating in the completed proposal. Translational Bioinformatics has been consistently one of the most popular courses our department offers. Each year I have between 20 and 30 students enrolled with most of the students coming from outside my home department.