[Editor’s Note: Kresten Lindorff-Larsen, Ph.D., a Professor at the Linderstrøm-Lang Centre in the Department of Biology at the University of Copenhagen, will be one of the speakers at the upcoming Mutational Scanning Symposium, May 21 through 23, in Barcelona. In this Q&A, Dr. Lindorff-Larsen discusses his upcoming presentation and his background.]
What is the subject of your presentation at MSS 2025?
I will be talking about the interface between variant effect prediction and interpretation. My colleagues and I are very interested in not just predicting variant effects, but also understanding molecular mechanisms of how they exert their functions. So, I will discuss work on computational models trying to assess — at the proteome level — which missense variants cause loss of function, and by which mechanism. In our initial work we have focused on separating variants that cause loss of function via protein instability and degradation from others that may more directly for example affect intermolecular interactions. We find that about 50 percent of disease-causing missense variants cause loss of protein stability and degradation. But then, what can we say about the other 50 percent? Is there something systematic about those? We are also very interested in understanding how proteins are degraded, including which sequences are recognized by the cell’s protein quality control system; we have been mapping these experimentally and training models to predict the rest.
Some of these points are discussed in a preprint with my colleagues from the University of Copenhagen, Matteo Cagiada and Nicolas Jonsson, “Decoding molecular mechanisms for loss of function variants in the human proteome.”
Please discuss your scientific background.
I originally studied biochemistry and had among other things been working on protein folding for some time. I started working in this area when my mother — a medical doctor — was working with people who had genetic forms of colon cancer so-called Lynch Syndrome. She was caring for patients some of whom had variants of uncertain significance, and she said to me, “Kresten, you know a lot about proteins, couldn’t you help figure this out?” I initially said, “No,” because I thought I couldn’t. But she kept insisting, so I eventually caved in.
So, I started working on this with that goal in mind, using our understanding of protein folding and stability. Very quickly, I was drawn into meetings where variants in real people were discussed. It was extremely inspiring and extremely intimidating. We publish many papers and we put out lots of results. But it also hits on a personal level, raising questions, “How should we treat this person?” And in computational models, “What would they suggest?” Answering those is really useful and makes you think differently about the way you do this research.
Around the same time, I started talking to my good colleague at the University of Copenhagen, Rasmus Hartmann-Petersen, who had been working on protein homeostasis and degradation for many years. At some point, we got together to see whether there was some overlap. Others had just shown that one of the Lynch Syndrome proteins was degraded in an interesting way, and there was our overlap. We started working together on this example, trying to understand which missense variants cause loss of function via simply protein degradation. And we sought to shed some light onto how this occurs, by combining experiments, biophysics, and statistical models of protein sequences.
Can you elaborate on the integration of bench science experiments and computational modeling?
One of the great things about this field is the tight integration between computational and experimental studies. As a field, we are using experiments to benchmark and parameterize our computational models, but also use computation to help drive experiments. My group does computational work and we work closely together with experimental collaborators. In this way, we can study a set of problems in the lab and learn general rules that we can then apply across a much broader set of proteins and systems that we have not studied experimentally.
'We are using experiments to benchmark and parameterize our computational models, but also use computation to help drive experiments.'
My view — and this may be a little simplistic — is that the best computational predictions of missense variant effects work really well, but generally at the cost of us not knowing much about the molecular mechanisms captured by the computational models. If you use this as a starting point for understanding disease mechanisms, potentially for developing therapeutics, then we think it will be fruitful to supplement these “mechanism-agnostic” models with a more mechanistic view so that we can then say, “Here are these great predictions. Can we try to get some mechanisms for what these variants are doing?” We call it doing proteome-wide biophysics.
Why are you participating in the Mutational Scanning Symposium this year?
This conference reflects a field of scientific study that — to a very large extent — has been driven both by technology and science. The meeting enables participants who recognize the importance of this technology to see new developments, and discuss what it might be used for.
At other conferences, I often come away thinking, “These are the things we learned, but not how did we learn those things?” Both are critical on the experimental and computational sides. The meeting covers aspects from molecular biophysics to patient data, and seeing this arc is a really useful reminder of why we do this. The Mutational Scanning Symposium really tries to bridge across this entire scale.
VISIT HERE FOR MORE INFORMATION ON THE SYMPOSIUM
