Ezequiel A. Galpern

biorxiv.orgBlending physics-based and inverse folding models to disentangle variant effects on stability and functionProtein sequences are constrained not only by the need to fold into stable structures, but also by specific functional requirements imposed by natural selection. Yet predictions of how amino-acid chan...

1/ New preprint! with @solersanchisx.bsky.social @cwjpugh.bsky.social @federicobilleci.bsky.social @jonnyfrazer.bsky.social and Mafalda Dias, we introduce a scalable framework to improve stability prediction and separate folding from functional constraints. www.biorxiv.org/content/10.6...

sciencedirect.comDark energy: the cost of function in protein evolutionThe evolutionary fate of proteins is driven by both folding stability and biological function, dual constraints that often conflict, creating frustrat…

"Dark energy: the cost of function in protein evolution" is now published in Current Opinion in Structural Biology! @diegulise.bsky.social @nachoquique.bsky.social link: www.sciencedirect.com/science/arti...

Ezequiel A. Galpern@eag91.bsky.social

1/2 🚀 We just uploaded our new review to arXiv: “Dark energy: the cost of function in protein evolution” With Federico Caamaño, @nachoquique.bsky.social & @diegulise.bsky.social 🔗 arxiv.org/abs/2608.27711

arxiv.orgDark energy: the cost of function in protein evolutionThe evolutionary fate of proteins is driven by both folding stability and biological function, dual constraints that often conflict, creating frustration and imposing functional costs beyond stability...

chemrxiv.orgFrom sequences and structures to fitness, folding and function: challenges in the age of AI | ChemRxivDeep-learning models have transformed our ability to predict the phenotypic effects of sequence perturbations. Protein language models (pLMs) score the evolutionary propensity of any amino-acid substi...

🧬 What do deep-learning models for proteins actually learn? Our new review looks at how model predictions relate to fitness, folding stability and function. With @cwjpugh.bsky.social, Mafalda Dias & @jonnyfrazer.bsky.social 🔗 chemrxiv.org/doi/full/10....

📊 Bigger is not necessarily better. Across ProteinGym benchmarks, scaling protein language models does not consistently improve zero-shot fitness prediction. We discuss why model likelihood is not the same as fitness, and what may be limiting performance.

Atlas of Variant Effects Alliance

Inverse-folding models have their own limitations: they can read functional conservation as a stability constraint. That matters when trying to disentangle folding from function. 🔧 We review these challenges and the emerging strategies to mitigate them.

2/2 🧬 Proteins need to fold, but they also need to function, and these requirements do not always agree. ⚖️ We review experimental and computational approaches that disentangle evolutionary constraints from folding stability, including our dark energy framework.

Ezequiel A. Galpern reposted

@varianteffect.bsky.social

10 Aug

@eag91.bsky.social was recently the 100th speaker of our Variant Effects Seminar Series. The work he presented has just been published as a preprint. Learn more: tinyurl.com/mrxf5zxv @solersanchisx.bsky.social, @cwjpugh.bsky.social, @federicobilleci.bsky.social, @jonnyfrazer.bsky.social

lnkd.inBlending physics-based models improves protein folding stability prediction | Ezequiel Galpern posted on the topic | LinkedIn🎉 New preprint: “Blending physics-based and inverse folding models to disentangle variant effects on stability and function” ➡️ Read it on https://lnkd.in/dCDw-3nb Together with Xavier Soler Sanchis...

github.comGitHub - DiasFrazerGroup/blend_and_disentangleContribute to DiasFrazerGroup/blend_and_disentangle development by creating an account on GitHub.

8 Aug

🚨 Paid MSc research opportunity at CRG Barcelona! With Mafalda Dias and @jonnyfrazer.bsky.social, we’re hosting a 4-month MSc project: 🧬 “Folding versus function: disentangling the molecular mechanisms of missense variants in gene-trait associations” More info 👇 www.crg.eu/en/content/t...

6/ We also released the full pipeline: • blend AWSEM and PottsMPNN • compute folding and dark energy for all single-site variants • visualise heatmaps and 3D structures • reproduce the paper figures github.com/DiasFrazerGr...

colab.research.google.comGoogle Colab

7/ Also you can try it directly in Google Colab: colab.research.google.com/github/DiasF... Feedback is very welcome!

4/ We then combine the resulting folding-energy estimates with evolutionary scores from a protein language model. This decomposes variant effects into: • folding energy • “dark energy,” capturing evolutionary constraints beyond folding

5/ The decomposition reveals distinct disease mechanisms: Gain-of-function and oncogenic variants often preserve stability but strongly perturb functional constraints, whereas many loss-of-function and tumour-suppressor variants are predominantly destabilizing.

2/ Our starting point: inverse-folding models predict stability changes surprisingly well, but they can confuse functional conservation with structural stability. We blend them with the physics-based AWSEM potential to combine complementary signals.

Atlas of Variant Effects Alliance

3/ The AWSEM–IF blends improve global ΔΔG prediction and reduce the systematic bias of inverse-folding models at functional sites.

Ezequiel A. Galpern reposted

@varianteffect.bsky.social

12 June

At last week's Variant Effects Seminar (VESS), @eag91.bsky.social discussed disentangling the biophysical effect of missense variants using protein dark energy. If you missed it, you can now watch it on our YouTube channel at a time to suit you. ⏯️ youtu.be/je5XosFTLPY

youtu.beDisentangling the biophysical effect of missense variants using protein dark energy - E. A. GalpernYouTube video by Variant Effects

Ezequiel A. Galpern reposted

Jonathan Frazer

@jonnyfrazer.bsky.social

1 June

Mafalda has been shortlisted for the 2026 Premio Vanguardia de la Ciencia, one of the more visible science prizes in Spain. Her interview: www.lavanguardia.com/vanguardia-d... ...and we would be very grateful if you could vote here: www.lavanguardia.com/vanguardia-d...

lavanguardia.comLa IA mejora el estudio de las enfermedades rarasVota a las finalistas del premio Vanguardia de la Ciencia

Ezequiel A. Galpern reposted

Atlas of Variant Effects Alliance

@varianteffect.bsky.social

20 May

Our Variant Effect Seminar Series is marking a milestone! Join us on Tuesday, June 2, for our 100th speaker, Ezequiel Galpern. Plus, special VESS guests, as we celebrate and look forward to the next 100 speakers! Join here: www.varianteffect.org/seminar-seri... #AtlasOfVariantEffects

nature.comAn information-theoretic argument for the restriction of the current biological alphabets to 4 nucleotides and 20 amino acids - Scientific ReportsLife as we know it is based on foldable biopolymers encoded with just 4 nucleotides or 20 amino acids. Evolution of these biopolymers requires effective and fast search of both the conformational spac...

1 Apr

Why are there 20 amino acids and 4 nucleotides? Combining Energy Landscape and Molecular Information theories provides constraints to the alphabet size of an evolving biopolymer, given its physico-chemical properties... Read more in our new article: www.nature.com/articles/s41...

@nachoquique.bsky.social @diegulise.bsky.social

1 Apr

3 Mar

elixir-europe.org/events/3d-bi... Next week!

pnas.orgPNASProceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...

22 Jan

"Probing the Dark Energy in the Functional Protein Universe" is now published in @pnas.org 🎉 www.pnas.org/doi/10.1073/...

Ezequiel A. Galpern@eag91.bsky.social

❗New Preprint w/ C Bueno (‪@cabb99.bsky.social‬), IE Sánchez (@nachoquique.bsky.social‬), PG Wolynes & DU Ferreiro (@diegulise.bsky.social‬) We define the Protein Dark Energy and use it to localize and quantify the functional evolutionary constraints on natural proteins arxiv.org/abs/2508.08109

arxiv.orgProbing the Dark Energy in the Functional Protein UniverseWe show how to localize and quantify the functional evolutionary constraints on natural proteins. The method compares the perturbations caused by local sequence variants to the energetics of the prote...

6 Nov 2025

Our method for predicting folding dynamics/mechanisms from sequence info is now a book chapter! link.springer.com/book/9781071... @diegulise.bsky.social

Ezequiel A. Galpern reposted

Jonathan Frazer

@jonnyfrazer.bsky.social

23 Oct 2025

Applications are open for the @crg_eu PhD Programme! 20 fully funded positions — including one in our group through the Evolutionary Medical Genomics ITN. Join us to develop deep generative models of cross-species data to tackle open questions in disease genetics. www.crg.eu/en/content/t...

Centre de Regulació Genòmica (CRG)@crg.eu

Are you looking for a PhD? Join us in Barcelona! You'll dive into a community of >100 PhD students from 30 countries exploring the frontiers of biology. You can also join an online workshop on 6 November (15:00 CET) to learn how to find the right lab for you. More info: www.crg.eu/en/content/t...

Some sites have, on average over all the possible variants, high Dark Energy. In those cases, there is a function beyond protein folding affecting natural selection. This is clearly the case of Enzyme catalytic sites

colab.research.google.comGoogle Colab

12 Aug 2025

Dark Energy is not a relative score, but provides a common scale and has physical energy units. And it can be also computed even if ΔΔG is not available, using AWSEM force-field. You can compute it for your favorite PDB colab.research.google.com/github/eagalpern/colabs/blob/main/DarkEnergy.ipynb

The idea is simple. For a deleterious mutation, we expect the protein to get destabilized. But how much? Correlations between a ΔΔG DMS and the corresponding ESM-2 scores are not perfect. We define the difference between these two free energies as a ‘Dark Energy’.

authors.elsevier.comauthors.elsevier.com

Finally published! “Inferring protein folding mechanisms from natural sequence diversity” Open acces link: authors.elsevier.com/a/1lQvh1SPTB... @biophysj.bsky.social A new chapter of an amazing project w/ @diegulise.bsky.social and Ernesto Roman 1/n

Also, we show that both the stability and cooperativity changes induced by mutations can be computed directly using sequence-based evolutionary models. 6/n

colab.research.google.comGoogle Colab

And you can make folding mechanism predictions for your favorite protein! colab.research.google.com/github/eagal...

authors.elsevier.comauthors.elsevier.com

For 15 diverse protein families, we computed the folding mechanisms of hundreds of proteins by simulating an Ising chain of folding elements, or foldons. The energetics is determined by each amino acid sequence. 4/n

Protein topology imposes limits on the variability of folding cooperativity within a family. While most beta and alpha/beta structures exhibit only a few possible mechanisms despite high sequence diversity, alpha topologies allow for diverse folding scenarios. 5/n

authors.elsevier.comauthors.elsevier.com

We know that closely related proteins usually share similar three-dimensional structures. But differences in their amino acid sequences can lead to distinct folding mechanisms, enabling the natural evolution of diverse biological functions. 2/n