How will neuroscience training need to change in the future?

Neuroscience is changing rapidly, especially with the advent of new artificial-intelligence tools. The field has become so diverse that no one can reasonably expect to be an expert in everything. Given this broad scope, how can neuroscience training programs best meet the needs of students and the field?

We polled our readers and contributors and engaged a market-research firm to conduct structured interviews with senior scientists from leading neuroscience programs around the world.

Respondents widely agreed that future neuroscientists will need stronger foundations in mathematics, statistics, computer science and machine learning. At the same time, researchers are concerned that students are overly focused on generating data and applying AI rather than thoughtfully analyzing results. Renewed emphasis on critical thinking and experimental design will be important to counter this trend.

Many respondents noted that neuroscience doctoral programs have become extremely competitive, so that only applicants with extensive experience are accepted. That limits opportunities for students from disadvantaged backgrounds, a situation that is likely to worsen with further funding cuts.

“We are not teaching enough quantitative computational skills. I see this acutely, at universities, among my collaborators, and [where] I teach in summer school. That’s why I do YouTube, in an attempt to provide training resources for students who want more computation but do not find it in the curriculum. There’s a large fraction of people that get to the postdoc or graduate student level and don’t know how to code or model—you don’t have to be fluent, but you need to be able to talk to a collaborator. These are large programs with serious curriculums that are woefully low.”
—Bing Wen Brunton, professor of biology, University of Washington

“Students need to become Keplers! They need to learn how to devise mechanistic, quantitative hypotheses to explain complex data. Nobody teaches them how to do this. Laboratories in virtually all branches of neuroscience are swamped by data. At most, 10 percent of data is used towards publications, a colossal waste. This is because typical biology students and faculty don’t know how to make sense of vast amounts of complex data. Stopgap solutions include black-box packages that do SVD, PCA, ICA, NMF, UMAP, etc. But students don’t know the limitations of these packages, potential artefacts, or what the results can/not tell us. As a result, many publications contain flawed inferences. The solution is to train students in basics like linear algebra, geometry, topology and numerical techniques, so that they can devise novel analysis tools appropriate for the questions at hand, not just use the packages. This will not only avoid flawed inferences and use the vast amounts of data, but [generate] qualitatively new insights about how the multiscale, large dimensional, nonstationary, highly nonlinear and distributed structure called the brain does its magic.”
—Mayank Mehta, professor of physics, astronomy, neurology and neurobiology, University of California, Los Angeles

“We need to find the right way to take advantage of AI. Programs and labs need to adapt. It will make a big difference.”
—Drew Robson, research group leader, systems neuroscience and neuroengineering, Max Planck Institute for Biological Cybernetics

“Data science will be increasingly important. It doesn’t matter if you do molecular neuroscience or system neuroscience. Most likely, you have to be a data scientist because we’re in the stage of just tons of data. We really need our trainees to be equipped with data science [skills], to be able to make the data they collect make sense. Support for scientific computing at the institution or university level is actually needed to equip our trainees with this skill set.”
—Lin Tian, scientific director, Max Planck Florida Institute for Neuroscience

“We need to educate a generation of neuroscientists who are able to think within computational frameworks while also understanding the fundamentals of current brain knowledge and the gaps that remain. The current generation often lacks either computational expertise or neuroscience foundations, reflecting the segregated nature of engineering versus neuroscience training backgrounds.”
—Shahab Bakhtiari, assistant professor of psychology, University of Montreal

“We need to change our ways. With AI and other things, the way students are trained as it applies to the lab is different than before. It’s difficult to get good students. I noticed a trend, that students are not really good at independently solving problems, of taking on the challenge. You need to try and fail, then keep going. You need to talk it through and work it out. But the new students seem to think the answers are always out there with Google or ChatGPT. If the answers are not, they don’t know what to do. They don’t want to try things. They see it as a waste of energy. We need independent exploration.”
—Martijn Cloos, associate professor of bioengineering, University of Queensland

“Traditionally, there’s been a barrier between theorists and experimentalists. That’s true in other fields besides neuroscience, but [in the piece I wrote for The Transmitter], I was making the argument that it benefits people to think both theoretically and experimentally. It’s good when the experimentalists are doing theory and the theorists are doing experiments; I’ve been trying to train people to do both and in my lab people do both. I think that that would be one place to change how we train people because currently there’s not a lot of impetus to do it that way. There’s a lot of division of labor.”
—Samuel Gershman, professor of psychology, Harvard University

“When I was educated, it was good enough for a scientist to be a specialist in the area and be really good at two or three experimental methods. And then you specialize, you can study what you’re interested in, and that’s good enough. I think when I look at the top neuroscientists today, they are very good experimentalists, but they are also very good in using computational tools in theory to make sense of their data, which are sometimes large-scale sequencing data, imaging data, genetic data. So, you can’t be limited anymore, I think. The top people, they sort of have a greater span in mastery of methods. And I think that is what training should aim to achieve.”
—Jorg Grandl, associate professor of neurobiology, Duke University

“I have noticed a surprising lack of focus on training students and even postdocs in the ‘science’ component of neuroscience. Neuroscience is a diverse field of research, and courses I have observed do a great job in exposing students to that diversity and providing resources for students to dive deeper into their particular areas of interest. However, this knowledge of neuroscience is not counterbalanced by the wisdom of how to apply that knowledge to answer new questions. Students want to rush headlong into collecting data without stopping to consider experimental design, let alone generating a testable hypothesis first. Perhaps it is a combination of the anxiety to publish or perish and methods of generating large-scale, mineable data that has led to this deemphasis of the scientific method and transformation of scientists into data generators as opposed to hypothesis generators. As we enter this new data-rich era, though, the pendulum is going to need to swing back.”
—Kyle Jenks, research scientist, Picower Institute for Learning and Memory, Massachusetts Institute of Technology

“I do think that graduate programs should increase efforts to make students in their first few years very familiar and competent in all major areas of neuroscience. It can be done. It can be done without extending the period of Ph.D. research.”
—Michael Stryker, professor of physiology, University of California, San Francisco

“There is a tension between the specialization that you need to really be able to understand and apply some of the advanced techniques that we have in the field and then not losing the forest for the trees. And that’s always very, very, very difficult where in an ideal world you would have your lab training, which involves using all these specialized techniques, and then it’s really down to your coursework to give you the bird’s eye view of why are you doing this? What are you trying to understand? What’s the overarching goal? And I think the best neuroscience training programs are doing that, where it’s almost like what you’re learning in class is philosophically tinged, and what you’re learning in lab is handiwork, and I think right now that’s the best model that I can conceive of. What’s really important is to help people understand that neuroscience is about the questions and not about the approaches.”
—Jan Wessel, Clement T. and Sylvia H. Hanson Family Chair and professor of psychological and brain sciences, University of Iowa

“The pressure is high, even in a successful, well-funded lab. You don’t know what will happen in one to two years. It will affect our mentorship ability, to allow many to have opportunities. Just the most motivated will be supported.”
—Luana Colloca, professor of pain and translational symptom science, University of Maryland School of Nursing

“We’re affiliated with Hopkins. The graduate program got cut in half this year—or more than half—with admissions, just due to uncertainty. This is true everywhere. The places that didn’t get oversubscribed and had to rescind acceptances anyways. So, we already took a pretty big hit as far as I know on this year’s training.”
—Joshua Dudman, senior group leader, Janelia Research Campus, Howard Hughes Medical Institute

“Funding will likely shrink. R01 grant renewals should be less common. More of the same is insufficient. Find a way to shrink gracefully. Smaller faculty. Fewer postdocs and students. Maximize leverage of AI. Communicate to the public what we do and why it is important to them.”
—Tim Harris, senior fellow, Janelia Research Campus, Howard Hughes Medical Institute

“Career prospects for young researchers are likely to be very difficult for the next few years, at least in the U.S. So they will need to be broadly trained and encouraged to develop transferrable skills that will enable them to pursue alternative careers. Those few who are very talented, completely committed to research and very tolerant of risk and uncertainty should keep doing what they are doing. Everyone else should be considering their plan B.”
—Charles Jennings, executive director, BWH Program for Interdisciplinary Neuroscience and the Romney Center for Neurologic Diseases, Harvard Medical School

“I think [U.S. funding and policy changes] are going to lead to a brain drain eventually. It has already started. It is harder now to attract foreign researchers either from China or from Europe to come and do a postdoc in the U.S. I think some established researchers in the U.S. are going to eventually throw up their hands and say, ‘I may have an opportunity in the U.K. or in Europe or in Singapore,’ and they will eventually have to leave if they really want to continue to do high-level research, especially in these topics that are controversial with the new administration.”
—Steven Proulx, group leader, University of Bern