Ten Simple Rules for Starting (and Sustaining) an Academic Data Science Initiative

Abstract

The past decade has seen an explosion of data science centers, institutes, and programs appearing across the U.S. as universities increasingly recognize the importance and promise of data science to university research and education. It has been, and continues to be, an exciting time. But there are systemic challenges faced by these initiatives in the context of the higher education system. Some, but not all, of these challenges center around funding. Campuses fortunate enough to receive initial funding, often as a result of philanthropy or private sector investment, have some measure of sustainability, especially if these funds are in the form of an endowment. However, at most smaller colleges and universities, or those without a lucrative alumnus or local industry investor, just getting started with very limited funding can be daunting. And yet, every school is facing the reality that to truly prepare their student body for the expectations of 21st century employers, they must find a way to incorporate core critical thinking and data-intensive skills into nearly every discipline. This call to action challenges traditional disciplinary silos and begs for new models of higher education. As data science initiatives attempt to breakdown old barriers, they each face a unique set of challenges due to their campus’ political, financial, and structural environments. Our consideration of these challenges and those we have faced directly ourselves lead us to recognize a set of global commonalities. We capture them here in the familiar Ten Simple Rules (TSR) format for simplicity, recognizing much more could be said.

A comprehensive survey of collaborative biostatistics units in academic health centers

“…The organizational structure of centralized collaborative biostatistics units in academic health centers (AHCs) in the United States and their important contributions to clinical and translational research have been widely discussed (Ciolino et al, 2021; Desai et al, 2022; Lee et al, 2018; Pomann et al, 2020; Rahbar et al, 2017; Rosenblum, 2012; Thiese et al, 2018; Welty et al, 2013). Establishing strong collaborative biostatistics units requires careful strategy, with important considerations including the unit's mission, along with its collaborator base, workforce structure, leadership, and funding model (Desai et al, 2022; Parker et al, 2021; Thiese et al, 2018; Welty et al, 2013). Desai et al (2022) outline four key elements that a data science collaborative unit should aim to incorporate (i) a collaborative philosophy; (ii) a funding mechanism that develops partnerships between the quantitative unit and a basic, clinical, or population science entity; (iii) a joint investment in the career development of faculty who practice data science; and (iv) training of faculty and staff in the practice of data and team science.…”

“…Pomann et al (2020) describe 12 steps that the collaborative process should follow, as well as a list of 16 competencies that collaborative biostatisticians should possess. Parker et al (2021) acknowledge the broad and multiple definitions of data science and the “explosion of data science centers” over the last 10 years and provide 10 tips for building a data science unit.…”

Is “bioinformatics” dead?

“…Data science—part computer science, part statistics, part information science, and part applied math—with its unique blend of methodologies and scientific cultures—doesn’t acquire meaning until put to practical use. Add to that the influence of the private sector, think DeepMind [ 3 ], and you have something new and groundbreaking. As an example, we describe ourselves as a “school without walls” — the “school” gives us autonomy in an academic institution, and the “without walls” means we don’t own anything but contribute to everything (or at least aim to).…”

“…Within our School of Data Science, we have given a great deal of thought to this question [ 3 ], focusing not so much on the definition itself, but rather on how we embody the meaning and culture of that definition in all aspects of teaching, research, and service to the community. As such, we have arrived at the 4+1 model of data science.…”