Pilar Cossio: Reconstructing RNA’s Shapeshifting Behavior

Flatiron Institute Senior Research Scientist Pilar Cossio builds tools to analyze experimental measurements and reveal how environmental changes alter the behavior of RNA molecules.

A photo of Pilar Cossio sitting in front of a shelf of math books. — Senior Research Scientist Pilar Cossio’s research focuses on developing mathematical and computational methods to characterize proteins’ structures and dynamics using cryo-electron microscopy, single-molecule spectroscopy and biomolecular simulations. Nicky Quamina-Woo/Simons Foundation

Inside your body are ribonucleic acid molecules that are hard at work serving as a bridge between your DNA’s permanent blueprint and the proteins that build, maintain and power your cells. These RNA molecules are highly flexible and dynamic. That’s useful for our bodies, but this changeability makes it tricky for traditional structural biology techniques to determine how RNA responds to environmental shifts.

Pilar Cossio, a senior research scientist at the Simons Foundation’s Flatiron Institute, develops computational tools to address this challenge. Her tools help analyze simulations and cryo-electron microscopy images to reveal how RNA’s structure adapts.

Cossio holds a B.S. in physics from the University of Antioquia in Colombia and a Ph.D. in the physics and chemistry of biological systems from the International School for Advanced Studies in Italy. We recently discussed her work at the intersection of math and biology.

How would you describe your work?

I have a joint appointment with the Flatiron Institute’s Center for Computational Mathematics (CCM) and Center for Computational Biology (CCB). I also work with colleagues at the Flatiron Institute’s Initiative for Computational Catalysis.

I build methods that enable biophysical inference. There are many unobserved dynamics that underlie the behavior of biological systems at the molecular level. For example, we do not yet have a good understanding of why RNA biomolecules fold the way they do, or what causes them to shift shape as they perform various tasks.

“Inference” means to determine these underlying dynamics as well as possible, with tools like molecular simulations and experimental data from high-powered microscopes.

My role is not to hypothesize inferences per se, but rather to build the tools that enable this work. That includes developing theories that can guide experiments, and proposing linkages between simulation and experimental data that drive interpretation and insight.

What drew you to this work?

I studied physics as an undergraduate, including astrophysics and particle physics. I also took a course in biology designed for physics students, which was my introduction to cell and molecular biology. That’s when I learned that cellular ribosomes translate messenger RNA sequences into the polypeptide chains that are the structural foundation of proteins. I also learned that RNA is a type of biomolecule, an organic compound essential to life.

Even today, I am amazed by how biomolecules function. And from the very start, I wondered: How does this work, from a physical perspective?

This has been the driving question of my work for 15 years now.

A photo of Pilar Cossio in her office, writing on her whiteboard table with a dry-erase marker. — Cossio works on a mathematical inference framework that compares cryo-EM images to simulated molecular structures. Nicky Quamina-Woo/Simons Foundation

What are some tools for biophysical inference?

One source of data is cryogenic electron microscopy (or cryo-EM), which uses electrons to visualize biomolecules flash-frozen in liquid ethane. This is our best tool for seeing biomolecules frozen in time.

But it’s not perfect. This data is super noisy. There are many uncontrolled variables — such as ice contamination or the fact that too much electron exposure causes radiation damage to the samples — that can make cryo-EM images challenging to analyze.

In addition to cryo-EM data, molecular dynamics simulations can also help us uncover hidden activity within biomolecules. These are simulations of what happens to the atoms of a three-dimensional biological system — a protein, say, or an RNA — in biological environments. The equations that govern the behavior come from Isaac Newton, specifically the second law of motion. As the simulation progresses, we see the proteins or RNA structure evolve and change shape.

My colleagues and I combine simulation and cryo-EM data to make biophysical inferences. In a way, my work is the glue that binds the work of other scientists together.

Why is RNA structure so hard to discern?

Compared to protein structures, many RNAs fold into 3D structures, but these are much more dynamic and quick-changing. For example, RNA helices can change direction and orientation very quickly.

We don’t yet have perfect tools for capturing these changes, as I noted with the challenges of cryo-EM. And even if we could perfectly visualize RNAs, we would still need to do more work to know why RNA evolved in the way that it does.

So this is a fascinating problem. And it’s more complex than with proteins, which, in comparison to RNA, have stable structures. In 2024, the developers of the machine learning program AlphaFold — which effectively predicts protein structure based on an amino acid sequence — won the Nobel Prize in Chemistry. And that was fantastic!

Given that protein structure is now at least somewhat understood, I wanted to focus on the thornier challenge of RNA structural ensembles.

A photo of a close-up of Pilar Cossio’s handwriting on her whiteboard table with a dry-erase marker. — Nicky Quamina-Woo/Simons Foundation

Please describe an experiment along those lines.

The Simons Foundation has an ongoing partnership with the New York Structural Biology Center, which has an amazing collection of cryo-EM microscopes.

We’re working with many copies of the same biomolecule, freezing and imaging them with an electron microscope. In these images, both the RNA helices’ orientation (the projection direction) and conformation (whether the helices are open or closed) are unknown. And we’ve placed these identical copies into a solvent with differing concentrations of ions (such as magnesium) to mimic what could happen in a cell.

Statistical analysis of the cryo-EM images enables us to make inferences about how the magnesium buffer induces changes to conformation. These findings enhance our understanding of how these RNAs adopt distinct folds in response to changes in the environment. We will publish this data soon. For this project, I am working closely with a colleague, Steve Bonilla, at Rockefeller University.

What are your thoughts about working at the Flatiron Institute?

I’ve been at the Flatiron since July 2021. We have a lot of freedom to pursue interesting projects; it’s a fantastic environment.

One program I’d like to highlight is CCBx, a satellite initiative of the CCB. One benefit of CCBx is very practical: It provides funding for theorists within the CCB to validate their theoretical models through experiments, which I did for the project at the New York Structural Biology Center with the help of a talented experimental postdoc.

In many cases, an experimentalist designs and runs an experiment from start to finish and may only talk to a theoretician if they get unexpected or puzzling results. But theory can inform experimental design if it’s considered from the start, and foster a synergistic relationship. That’s what I do within CCBx. I appreciate that it highlights theory’s role, and I look forward to continuing my work at Flatiron for years to come.