Changing Workforce Development Needs

What are some of the most pressing challenges in workforce development, Bollenbach asked the panelists, particularly related to data science, AI, and computational biology? John Balchunas, the workforce director at the National Institute for Innovation in Manufacturing Biopharmaceuticals (NIIMBL), agreed that there is a need for additional manufacturing and quality talent. Although the fields of biopharmaceuticals and regenerative medicine are distinct, the workforce needs in these fields are similar, he said, and there are two challenges related to the changing workforce development needs that have been spurred by the proliferation of systems thinking. First, there is a need for re-training, or upskilling, the existing workforce so that they are able to interact with and respond to data, rather than routinely performing the same tasks each day. Given the interdisciplinary nature of this type of work, scientists should be able to interact with various stakeholders and operate in different modalities as needed, Balchunas said. Second, he continued, industries need to find ways to attract skilled, multidisciplinary talent. New approaches are needed to attract Ph.D.s, undergraduate students, and others to the fields of regenerative medicine and biopharmaceuticals—industries that are often in direct competition with large technology and software development companies that may appear more desirable for talented students preparing to enter the workforce.

Addressing Recruitment Challenges in the Field of Regenerative Medicine

Many talented students who would be well suited for work in the field of regenerative medicine are offered high-paying jobs in other industries even before completing their educational programs, said Alison Gammie, the director of training, workforce development, and diversity at the National Institutes of Health's (NIH's) National Institute of General Medical Sciences. The allure of a high-paying job is often strong, especially for students who have accumulated debt to pay for their education. Creative solutions should be developed to help workforce entrants manage their student debt burdens, she suggested. This issue also has a diversity dimension, she added, because underrepresented minorities carry a disproportionately high debt load (Addo et al., 2016).

Dawn Tilbury, the assistant director of engineering at the National Science Foundation (NSF), suggested that in addition to addressing the student debt load, it would be helpful to educate students in relevant areas of study about the importance of work being conducted in the regenerative medicine and biopharmaceutical fields. Students might be swayed if they realize that their contributions in these fields may have great social impact and may help save lives. The appeal of making an impact may offset the appeal of a high-paying job at a large technology firm. She added that students from underrepresented groups tend to have a greater affinity for jobs that have a tangible impact on people's lives. Manufacturing may have less allure than other technological fields, but it can offer workers high-impact, meaningful careers, Balchunas said. The industry should more clearly communicate this aspect of working in regenerative medicine, he added.

Training and Education Considerations in Workforce Development

Robert Zambon, the senior director of data strategy and external innovation at Johnson & Johnson, highlighted the need to recruit data scientists, analysts, and machine learning experts to the biopharmaceutical fields. Conversely, students in biology, chemistry, and other related scientific fields should be actively recruited into data science programs while still at university, if they show an interest. Students who study data science, AI, and machine learning tend to consider working either in the high-technology sector or in the biopharmaceutical industry—and view them as two distinct fields. Continuing to provide information about the different career options available to those who are currently in higher education—as well as those who are preparing to enter the workforce—would help to shift the balance and shape the workforce needed in the biopharmaceutical and regenerative medicine fields. Zambon also cautioned against focusing exclusively on trying to create polymaths who can do everything; instead, a better approach might be to focus on building teams in which different individuals have training in different areas of expertise.

Balchunas agreed that individuals need not be polymaths but said they should have a breadth of skills and competencies about the different disciplines they will act with in the field. The question, he said, is what foundational level of knowledge is needed for people working at the various levels within the biopharmaceutical and regenerative medicine industries. Providing a wide variety of skills and competencies as part of training may require a paradigm shift in the way the workforce is educated, he suggested. Tilbury drew an analogy with people who are able to speak not just their native tongue but other languages as well: students preparing to enter the workforce in these industries should be trained to speak a few words of the languages of other disciplines (e.g., chemistry, biology, data analytics, computer science) in order to remove the language barrier that can impede progress and collaborative problem solving in interdisciplinary fields. NSF is interested in exploring challenges related to interactions between disciplines to solve specific challenges, she added. In this type of “convergence research,” multiple disciplines come together to learn each other's languages, tools, and data methods; this approach even offers the potential to create new disciplines. Bioinformatics, for example, is a relatively new discipline that emerged from the intersection of biology and information theory.

Promoting Interdisciplinary Competency

Are there ways, Bollenbach asked, to help biologists cultivate a greater appreciation and understanding of the physical sciences? Teams, Zambon suggested, should be built in a way that allows experts in different areas—such as an expert physicist and a biologist—to coordinate and work together toward respective goals that are similar, but not identical. When this type of collaboration is successful, the experts will often develop a common language through which they can understand each other's work, even if they are sometimes using the same term to represent completely different concepts in their respective disciplines. Developing these types of teams can bring together experts in data science, mathematics, computer science, biology, and genomics. Zambon described the result as a continuum; despite the variation in expertise and experience across this type of team, individuals start to come together naturally to fill gaps and build out new capabilities. In developing the workforce, he added, it is important to engage with undergraduate and graduate students to provide them with exposure to work in the field early on in their career development trajectories, for example, through fellowship programs and opportunities to rotate through different internal and external rotations. The strategy of bringing together a group of individuals with different areas of expertise to collaborate in an interdisciplinary project setting is relatively common in academic settings, Balchunas said. For instance, he runs a professional development program at the Biomanufacturing Training and Education Center at North Carolina State University¹ that assembles those studying engineering as well as different types of sciences at various levels to work together on specific projects. No matter what their educational background, the participants are learning all of the same content and taking the same approach, which, he said, has been effective in this program and at other universities with similar programs.

Bringing Systems Thinking to Regenerative Medicine Through the Workforce

How, Bollenbach asked, can industry and regulators help prepare the workforce to bring a systems thinking approach to regenerative medicine? From an industry perspective, Zambon replied, frequent engagement can help catalyze the move toward novel approaches and using data across the board in regenerative medicine. Regular networking can also help build relationships among various actors within regulatory agencies as new regulatory programs are instituted in different agencies. For instance, building relationships with regulators at the Food and Drug Administration (FDA) can ensure that they are familiar with the new products under development. FDA works extensively with various entities to help develop new plans and review activities on an ongoing basis, which can help identify routes to evolve. This provides regulatory agencies with insight into the field and helps coordinate industry-wide efforts. There could be a role for NSF engineering research centers and manufacturing innovation institutes (e.g., NIIMBL, BioFabUSA) to interact with FDA through a public–private partnership to transfer this knowledge, Bollenbach suggested. Tilbury agreed, emphasizing the importance of early engagement and partnership between industry and regulatory bodies. Inviting representatives from regulatory agencies such as FDA and from manufacturing innovation institutes to participate in NSF's engineering research centers (ERCs) could be helpful, she suggested, so that they can follow research from its early stages to understand how the field will be moving forward. When NSF funds a phase I Small Business Innovation Research (SBIR) or Small Business Technology Transfer (STTR), FDA representatives are invited to the kickoff meeting with the researchers, which helps keep the regulatory environment in mind from the outset, she said.

Developing Roadmaps for Workforce Development

Asked whether there are training roadmaps being developed, Tilbury replied that some of NSF's ERCs develop curricula for students and training material for the next generation of the workforce. It would be helpful to engage with ERCs and other large centers that are simultaneously educating students and doing research to find out which types of skills they are seeking in their research students. NIIMBL has funded projects in education and workforce development that have developed curricula in areas such as data modeling, automation, and advanced analytics, Balchunas said. He said he was unaware of efforts to look specifically at workforce needs related to data analytics and systems, but he suggested that roadmaps developed for other industries could be adapted and applied. Gammie said that there are training programs available in this area. As a requirement of their NIH training grants, these programs are required to have specific, measurable goals in terms of students' outcomes and the skill set required to transition to the next level of the program. These programs are developed with a strong understanding of the career pathways available to the scientists they are training, which have overlap with a lot of the areas relevant to the workshop.

Bollenbach said that NSF recently funded an effort at the Georgia Institute of Technology² to develop a roadmap to look specifically at these education and workforce development challenges. This is important, he added, because while education and workforce development are very much a part of the work being done by BioFabUSA, NIIMBL, and the other manufacturing innovation institutes, their mandates are broad enough that they are often unable to dive in as deep as they would like on certain components, such as data science.

Janssen and Johnson & Johnson have established data science academies that focus not only on training and developing staff, but also on educating executives, vice presidents, and corporate-suite officers about new capabilities, including systems thinking, that are emerging, Zambon said. Leadership also need to understand the common language and overall value of the new methodologies and approaches as they transition from the traditional approaches of past decades. Bollenbach agreed that members of leadership need to be educated about these novel approaches so that they can explain to investors why additional research and development is needed up front in product development to select sufficient and appropriate data prior to initiating clinical trials.

Strategies to Strengthen the Workforce

What steps can be taken within the next 3 years, Bollenbach asked, to strengthen and prepare the regenerative medicine workforce? There is a need for collaboration and synergies across various initiatives and stakeholders (e.g., NSF, BioFabUSA, NIIMBL, academia) to reduce the potential for redundancy by sharing ideas more broadly, Balchunas said. Tilbury suggested that the concept of “workforce” may be too vague. The concept encompasses Ph.D. students considering their career options, members of the existing workforce in need of new training, technicians, and others at different levels of the workforce with various types of expertise. She agreed with Balchunas's point about synergies, suggesting that there should be an inventory of what is currently happening in order to be able to identify specific gaps, needs, and opportunities in education and development for different cadres within the broader workforce. NSF would be willing to host a workshop to bring stakeholders together for such a discussion, she said. Zambon suggested raising awareness about how systems thinking, data science, analytics, AI, machine learning, and related fields affect people's everyday lives. Within the industry, internal data-science symposia could bring together an entire organization to show people the impact of data science, systems thinking, and other new approaches. These kinds of events can pique the interests of younger staff members and encourage them to seek training and find ways to integrate these new approaches into their own work on a day-to-day basis.

Bollenbach concluded the session by raising several open questions for future consideration. How do these topics affect broader workforce training? How should training in the regulatory science sector be carried out in order to take data science and systems thinking approaches into account? If AI is augmenting automation down the road, how will that affect the way quality control is done? In this case, quality control will not be carried out through chemistry, but rather by watching sensor data. How will validation be done, and will that change?

Theoretical Approaches to Large-Scale Data

The first session, Plant said, had focused on abstract ideas, theoretical models, and ways of thinking about data, information theory, landscape and gene network models, dynamic heterogeneity, and other high-level concepts. Access to large amounts of molecular-scale data does not guarantee the ability to predict what a biological system will do, she said; in addition, the data must be analyzed within an appropriate theoretical construct. Theoretical approaches for examining complex biological processes at the systems level largely involve assessing the probabilistic nature of biological outcomes and understanding how molecular components work together to produce the emergent biological characteristics of the system. The importance of emergent properties means that reliance on reductionist research approaches is no longer sufficient. It will be important to develop theoretical frameworks that help interpret the results of today's powerful omics tools and make it possible to understand how to use molecular data to predict emergent properties. Roy said that the discussion on potential landscapes and high probability attractor states suggests a guide for how to control manufacturing and differentiation processes. However, he said that realizing this promise of identifying the desirable cell population landscapes (e.g., the ideal cells to inject or use to make a tissue) will require integrating patient-level data into the systems-level analyses.

Challenges of Systems-Level Approaches

During the second session, Plant said, panelists converged on the ideas that a systems approach is critical for developing and manufacturing clinically successful products but that there are substantial challenges involved in implementing that approach. For instance, substantial resources will be needed to collect and analyze the volume of systems-scale data needed to identify emergent properties and develop new models.

Cross-Sector Collaborations for Developing and Sharing Datasets

The third session focused on how collaborative work across industry and academia through consortia can be applied to develop large data-sets—including real-world data—that can be analyzed using sophisticated methods, such as AI. Plant said that the presentations of examples of large-scale data-sharing efforts ongoing in the drug discovery and health care spaces demonstrate that it may be possible to achieve sufficient volume and scale of data to provide insight into complex patient profiles. These efforts suggest that data sharing at an even greater scale has the potential to provide a high-variable space in which to query patient response to complex products. Developing a common language, establishing cooperative strategies for collecting and sharing data, and making a strong business case for aggregating and sharing these large datasets could further facilitate coordinated efforts to identify major effects and emergent properties, which would in turn allow follow-up research on specific systems to identify specific targets. Plant noted that some presenters had indicated that regulators are increasingly interested and embedded in all of these activities. Roy added that patient data collected through consortia could be used to better understand how therapies affect patients at individual and group levels. Additionally, he said, active participation by regulators in this area will also be beneficial by reducing costs, reducing risks for industry, and accelerating product development.

Use of Data Modeling to Reduce Data Dimension to Key Variables

The fourth session explored the variety of model types and methods that can be applied within systems thinking approaches. A systems approach can help home in on those analytes and relationships that provide the most important predictive value, but identifying the most important contributors and their relationships requires analytical approaches that are designed to collapse state space. Plant said that these models go beyond simple cluster analysis by addressing matrix factorization in order to better understand how cell populations transition between phenotypes and by identifying cell dynamics, heterogeneity, and other characteristics. She added that there are significant modeling challenges to collapsing the molecular detail space to permit the identification of the major interactions that drive phenotype.

The fifth session, as Plant recounted, featured examples of how industry is using these types of systems-level modeling approaches, including a hierarchy of models to explore manufacturing processes. The negative consequences—both for patients and for industry—of failing to adopt a systems thinking approach in regenerative medicine were well established throughout the workshop, she said.

Cross-Disciplinary Training of the Workforce

The sixth session explored challenges related to recruiting talented students with computational skills into the regenerative medicine space when there are other lucrative career paths available to them. Plant highlighted the discussion around building teams containing a diversity of expertise rather than seeking out individuals who can “do everything.”