My collaborator Erik’s research group, my professional home this summer, operates out of the Science for Life Laboratory, a five year-old center for molecular biosciences sponsored by four Swedish universities. Erik holds faculty positions at two of the member institutions, Stockholm University and the Royal Institute of Technology; the lab is housed at a third, the Karolinska Institute, on the northwest end of Stockholm City. The striking cylindrical facility—a block north of the iconic Nobel Lecture hall, Aula Medica—is one of several construction projects reshaping the Karolinska campus and the Stockholm research community.
A decade earlier, an institute called Science for Life might have been filled with greenhouses, petri dishes, and rat mazes; and it does house a few. But our floor, among several in the complex, looks more like a tech startup, with clusters of standing desks, wide-screen displays, whiteboards and breakout rooms. Built in 2010, SciLife dedicated substantial space to data science, including bioinformatics, high-resolution imaging, and Erik’s specialty, molecular dynamics. With his help, I aim to simulate in silico some of the molecular processes we observe in vitro in the lab—and to integrate computational and functional techniques in my own research workflow. Some thoughts on the human ecology of this transition below.
Power tools: Erik’s group steers the development of GROMACS, a free software package for simulating atomic-scale motions of macromolecules. Since its inception in the 1990s, GROMACS has been applied to an enormous range of protein, lipid, and nucleic acid systems; among other things, it provides the predominant base code for the popular folding@home distributed disease research project. One PhD researcher in Erik’s group works full-time to maintain and develop GROMACS; other team members work on various biomedical applications, including voltage sensing, structure determination, and my area, receptor modulation. Erik recently hired another PhD scientist to establish a small electrophysiology lab two floors down; I hope to make myself useful in scaling up their efforts.
Dream team: In addition to the GROMACS and wet-lab leads, Erik’s current group includes a half-dozen PhD students, plus two postdoctoral fellows combining molecular dynamics with other computational methods. An Associate Professor in the group supervises three more PhD students on accelerating simulations. Another full-time PhD incorporates X-ray scattering to visualize fine motions of membrane proteins. An industry colleague from a local biomedical company is on-site to simulate interesting physiological targets. And of course, we all work at the mercy of the group’s systems administrator.
Different world: The SciLife facility feels a little space-age compared to my home institution. Skidmore has excellent facilities, but focuses on providing undergraduates with a liberal arts education, including but not specializing in technical fields like computational chemistry. My lab back home is populated by about a dozen undergraduates (though usually no more than six at a time) sharing stations for buffer preparation, molecular biology, microbiology, injecting and recording. My students so far have only dabbled in modeling, working on laptops in the student lounge or my office. Incorporating more of a computational focus may require some changes.
Challenges ahead: Recruitment is less of a problem than I anticipated; our undergraduates recognize the increasing importance of computational research, and are probably better equipped than faculty at transitioning between technologies. Appropriate technical training can be a problem at a small liberal arts college, but I’ve found our students are more computer-savvy than even they believe. A bigger challenge has been resource equity. Most US college students own laptops, but their devices are often clunky hand-me-downs, weathered by constant travel between dorms and classrooms. This works fine for term papers and light data analysis, but anything more technical (specifications, installation, file management) can be challenging—and exacerbate inequalities we would rather help to level. Another valuable resource is space: computational work in a library or dorm room is perfectly feasible, but lacks the community of smart, interesting colleagues that makes lab work so engaging. Erik’s tech-company environment is an interesting alternative—though it would take some creativity to implement in our eternally strapped physical plant back home.
Open access: On a hopeful note, what appeals to me most about the digital revolution in biochemistry is the accessibility of many (if not all) relevant tools. To be sure, it still takes Erik’s team over a month to simulate just a few microseconds in the life of a fully solvated membrane protein; but many applications are lower-powered. Free packages like UCSF Chimera enable 3D visualization and limited editing of macromolecules on almost any laptop. Coarse-grained analytical approaches like principal component analysis and elastic network models—originally developed for problems in the visible world—can meaningfully approximate atomic motions. Several tools for sequence analysis, structure prediction, and docking/pocket analysis are available through free web servers.
In reach: Even the preparatory steps for intensive molecular dynamics simulations—generating topologies, solvating, energy minimization, equilibration—can be run locally prior to submitting to a cluster or supercomputer. And the increasing availability of low-cost computational resources through cloud services like Amazon and Google could enable students and other nonspecialists to conduct sophisticated analyses for a few dollars per hour. For undergraduates, the flexibility of computational research is also valuable: unlike culturing cells, students can set up simulations to run while they head to class, and analyze data any time afterwards without losing viability. Tools like these can open up research opportunities across disciplines, resources, locations, and schedules, if you know what to look for.
(The potential efficiencies of computational research were among the selling points for our department in securing a new tenure-track line, now open—I’d be thrilled to find a future collaborator. Get in touch!)
Big picture: Computational methods offer an enticing opportunity to efficiently connect reductionist experiments with a wider worldview. Whereas systems biology models processes through an entire organ, organism, or population, molecular dynamics models the motion of a macromolecule in its atomistic environment through time. Thus, rather than replacing wet-lab experiments, computational techniques (at least for now) augment the interpretation of functional data—hence the need for effective integration. And they may enable practical innovations: modern imaging techniques like cryo-electron microscopy have thrived on processing strategies equally applicable to face and sound recognition. With an ever-increasing selection of widely accessible tools, the real challenge may be choosing wisely—and knowing when and how to persevere, try something new, or ask the experts.