Catalyzing Multidisciplinary Frontier Science

A submission to the Astera Institute's Essay Competition (astera.org/essay-competition/)

Some scientific projects begin with a principal investigator (PI), a long-term vision, a lab, and a grant or two to fund it. Others begin with a casual conversation over lunch.

An example is a recent project I’m proud to have been a part of, which focused on quantifying how the rapid freezing process used in cryo-electron microscopy distorts the protein ensembles the method is meant to measure. This is a thorny problem at the intersection of structural biology, water and protein dynamics, and nonequilibrium thermodynamics. I could write way more about why this was an especially interesting and important problem to solve, but for the purpose of this post I’ll focus on the higher-level metascientific structure that enabled our success.

My involvement came about because I was visiting the Flatiron Institute to meet with collaborators on something completely different, and by chance ended up having lunch next to a Research Fellow there who was in the early stages of working on this project. He’d heard I worked on problems at the intersection of statistical inference and nonequilibrium thermodynamics, and asked whether I had any ideas for approaches to correct for statistical biases caused by the cooling process. I had a few ideas, and ended up being inducted into what became one of the most fun scientific collaborations I’ve worked on, and which I think will become a foundational pillar supporting future work using cryo-EM to measure protein structural ensembles.

This project was particularly unique in the breadth of specialists that were involved in bringing it to life. Making it happen required bringing together experts on cryo-EM, molecular dynamics simulations, Markov-state-model construction, water vitrification physics, structural biology, and nonequilibrium thermodynamics. This ultimately meant 5 postdoctoral researchers all making significant contributions (the final paper has 5 main figures, each revolving around the work of a different team member). This kind of truly multidisciplinary collaborative effort is pretty rare, so how did this happen?

1. Most of the scientists (including myself) doing the main research work for the project were independent postdoctoral fellows with their own no-strings-attached funding allowing them to work on whatever they felt was most interesting.

2. The Flatiron Institute catalyzed the collaboration by providing significant computational resources and funding to bring all the researchers involved together periodically at a single physical location.

3. A shared vision to put in the effort to create something really special even if this work may not be perfectly aligned with everyone’s broader career incentives.

These factors are the exception rather than the norm, and structural bottlenecks across the board make projects like this extremely difficult to both start and carry forward. Modern science is good at funding established labs working on long-term projects. What it is less good at is catalyzing the nucleation of new projects outside of established labs, especially when those projects require a team that is temporary, cross-institutional, and non-hierarchical. I believe we need to create new mechanisms to support project nucleation by non-PI scientists with rare, high-leverage expertise.

A mismatch of timescales makes project nucleation at the frontier a challenge
How are new ideas generated and funded in academia? The dominant pathway to generating and funding research ideas is that a PI sketches a ~5 year vision involving multiple connected projects and applies for funding in the form of research grants to make it happen. The 5-year timescale lines up both with typical grant durations and typical lengths of PhD programs, so it’s not surprising that this has become the dominant timescale. Interestingly, the 5-year timescale has also become central to more non-traditional research structures that have emerged in recent years, such as FROs and ARIA programmes.

I’m not here to argue against the long-term view of PI-led research programs, because I think this is an integral part of the scientific ecosystem. But it need not, and should not, be the only mechanism for generating and funding new ideas. On a shorter timescale than the 5-year research grant, the researchers at the scientific frontier are mostly transient, moving from place to place on temporary contracts of 1-3 years. This makes it impossible for them to take ownership of traditional multi-year grants, especially when these grants have application cycles of many months or more. Unfortunately, these transient specialists are exactly the people most likely to stumble across the major problems that will motivate future research directions.

Ideas generated through chance encounters by experts in different fields at different institutions are often lost to time, or relegated to a couple bullet points buried in a notebook, because existing research funding mechanisms are too slow and too large to spring into action before these scientists are forced to move on in their career path. Likewise, existing funding mechanisms are structured around supporting a single PI leading a lab, rather than a non-hierarchical team of experts from different disciplines.

Career incentives reward project ownership, not catalysis
Another key bottleneck in nucleating new projects is that in many cases the expertise needed to solve a given problem does not exist within any single lab. In principle, this is what collaboration is for: the scientist leading the project identifies an expertise gap, and reaches out to someone closer to that gap.

The difficulty is not usually identifying and contacting an expert collaborator in the relevant area. What’s much harder is convincing them to work on the project deeply enough for it to succeed. An outside expert with critical expertise who comes in and helps solve a specific problem will generally end up as a middle author on the eventual paper. Since (for better or for worse) authorship on papers is the main currency for credit in scientific research, with first and last author positions being by far the most valuable (although this is somewhat field-dependent), the incentive mismatch is clear: researchers with rare and useful skillsets may be genuinely excited by the project, recognize that it is important, and know that their contribution could make the whole thing possible. But from a pure career standpoint, they are often better off spending that same time on projects they lead themselves. The result is that science systematically under-incentivizes exactly the kind of expertise-sharing that many transformative multidisciplinary projects depend on.

How do we realign incentives so that researchers with rare, useful skillsets don’t have to choose between prioritizing their career and catalyzing transformational research that wouldn’t be possible without them?

Why are these problems particularly urgent now?
As the edifice of scientific knowledge grows taller, each researcher’s expertise occupies a narrower slice of the frontier. This is reaching a point where often even two researchers in the same subfield no longer speak the same scientific language. The most impactful research work succeeds by bringing in experts from different subject areas to cover a wider swath of the frontier. To address the systemic bottlenecks outlined above, I propose two new mechanisms to better catalyze the kind of multidisciplinary collaboration I’ve described: nucleation grants, and catalytic researchers.

Proposal #1: Nucleation Grants
To make starting collaborative research projects more accessible for scientists working at the frontier, I propose the mechanism of a nucleation grant. The goal of the nucleation grant would be to catalyze assembly of a multidisciplinary team of scientists across institutions who, combined, possess a unique combination of skills necessary to solve a problem at the frontier. They would be aimed at problems that require multiple complementary expertises, lack a natural single-lab home, and can be meaningfully de-risked in under a year.

Unlike more traditional research grants, these would be fast (rolling applications, with a turnaround time on the timescale of weeks rather than months), short (funding for 3-12 months to rapidly de-risk and evaluate a new idea), small (on the scale of $50,000-$200,000), and non-hierarchical (bringing together a team of equal collaborators rather than flowing to a single lead investigator). They would support a small team either working nearly full-time for a few months, or part-time for longer, to rapidly iterate on and evaluate the potential of a new idea. Perhaps even more so than explicit funding, an important part of the nucleation grant would be providing a physical place for the collaborators to come together (either for short periods of time or for longer bursts), along with travel support to bring people together and whatever resources are necessary to support the emerging scientific work. Such resources might include CPU/GPU time, credits for the latest and greatest AI tools, or physical laboratory space and core facilities.

A nucleation grant program would be successful if it led to the creation of new multidisciplinary teams that went on to do exceptional research work together. Something like this exists as of recently in the form of the National Institute for Theory and Mathematics in Biology’s Focused Research Group Awards program, but this only solves the physical location/travel portion of the problem, and is at present restricted in scope to only mathematical biology.

Proposal #2: Catalytic Researchers
To address the challenge of incentivizing collaboration and expertise sharing, I propose the framework of the catalytic researcher. This would be a new role for scientists with a rare and unique set of skills (e.g. expertise in a particular technique, method, or framework) that could be applied to help solve open problems across many different fields. Perhaps structured as a postdoctoral fellowship, or a grant supporting a staff scientist, these researchers would be embedded within a university or private institute with a renewable contract a couple years in length, and funds to support their activities, with a particular focus on travel to nucleate new collaborations.

The key difference between a catalytic researcher and a more traditional postdoctoral researcher, staff scientist, or junior professor, would be the criteria for success. While traditional researchers are evaluated on the degree to which they lead impactful projects to completion, the catalytic researcher would be evaluated on the quantity and impact of research outcomes that were made possible by their irreplaceable contribution. Such a researcher would be someone passionate about collaboration, happy to spend significant time chatting with other researchers looking for opportunities to catalyze progress in ongoing research efforts, both within, and especially outside, their institution. They’d work on a portfolio of different projects across a variety of fields, and play a key role in enabling someone else’s vision rather than being laser-focused on enabling their own.

As an aside, I think a successful catalytic researcher would be perfectly positioned either become an ARPA-type program director, or to start a BBN- or Frontier Research Contractor-type company focused on bringing the tools they’ve honed to new problems across academia, government, and industry.

Conclusion
The collaboration I described at the beginning of this post worked because several unusual conditions aligned: flexible funding and research freedom, available resources, physical co-location, and shared vision. In order to better support this kind of multidisciplinary frontier science, we need mechanisms designed specifically to lower the barriers to enabling these conditions. Nucleation grants would make it easier for new teams to form around an emerging problem. Catalytic researchers would facilitate key contributions from rare, high-leverage experts. Together, they would help modern science do something at which it currently struggles: turn promising conversations at the frontier into real research programs.