Valerio Piomponi

Molecular simulations and AI at @areasciencepark
Fromer PhD student at @bussilab.

[JCIM & JCTC Journals @JCIM_JCTC]

Evolutionary Constraints Guide AlphaFold2 in Predicting Alternative Conformations and Inform Rational Mutation Design #AlphaFold
https://t.co/UTjpwonYdk
@Piompons @albecazzaniga @AreaSciencePark
#JCIM Vol65 Issue18 #MachineLearning #DeepLearning

[Valerio Piomponi @Piompons]

Our study is now published on JCIM🎉

We expanded and refined the preprint thanks to the insightful feedback from reviewers!

paper: https://t.co/jWCkDipUZ1
code: https://t.co/XQb960wUxq

[Biology+AI Daily @BiologyAIDaily]

Evolutionary constraints guide AlphaFold2 prediction of alternative conformations and inform rational mutation design

1. This study introduces a refined method to guide AlphaFold2 toward predicting alternative protein conformations by clustering homologous sequences using MSA Transformer embeddings and agglomerative hierarchical clustering, enhancing interpretability and sampling efficiency.

2. The key insight is that AlphaFold2 predictions are strongly influenced by evolutionary signals embedded in the multiple sequence alignment (MSA), often favoring dominant coevolutionary patterns over the most thermodynamically stable structure.

3. The proposed clustering strategy generates more cohesive and informative sequence groups compared to density-based DBSCAN, enabling AlphaFold2 to sample high-confidence alternative states across diverse fold-switching proteins with minimal false positives.

4. By integrating Direct Coupling Analysis (DCA) into the clustered MSA pipeline, the method identifies co-evolving residue pairs specific to alternative conformations—providing a rational basis for mutation design to stabilize desired structural states.

5. The authors demonstrate the utility of their method on multiple proteins, including KaiB, RfaH, and proplasmepsin. In KaiB, a double mutation (E31H-P67E) was designed and validated via alchemical free energy calculations to stabilize an alternative conformation by 3.5 kcal/mol.

6. Principal Component Analysis of AlphaFold2's internal Evoformer representations reveals that structural transitions are encoded early in the model, particularly in pairwise residue interactions, underscoring the model’s latent capacity to encode multiple conformations.

7. The framework is not limited to fold-switching proteins—it also captures biologically relevant transitions in GPCRs and kinases, such as disulfide rearrangements in the Beta-1 adrenergic receptor and activation loop shifts in EGFR tyrosine kinase.

8. In EGFR, the method recovers the inactive state with high confidence and identifies a triple mutation (Y158C-H159L-K164N) that may favor this conformation by mitigating electrostatic repulsion and promoting close-contact residue interactions.

9. The strategy offers a computationally efficient and interpretable way to sample metastable protein states, making it a powerful tool for mechanistic structural biology and rational protein engineering.

10. By uncovering evolutionarily encoded conformational diversity, this work enhances the functional utility of AlphaFold2 and highlights the importance of coevolutionary analysis in guiding structural predictions and mutation design.

@Piompons @areasciencepark @albecazzaniga
💻Code: https://t.co/oyR7mnZEb8
📜Paper: https://t.co/O3EJLja6S6

#alphafold #proteinstructure #bioinformatics #coevolution #computationalbiology #mutationdesign #proteinengineering #foldswitching #AI4Science #structuralbiology

[Valerio Piomponi @Piompons]

Just started my migration to🦋, I hope to see all of you there very soon!

[Biology+AI Daily @BiologyAIDaily]

MDRefine: a Python package for refining Molecular Dynamics trajectories with experimental data

• MDRefine is a novel Python package that refines molecular dynamics (MD) simulation trajectories by integrating experimental data, improving alignment with observed molecular dynamics, particularly when force-field inaccuracies are present.

• This tool allows for combined refinement of ensembles, force fields, and forward models, making it highly versatile and effective in matching MD predictions with experimental observations.

• MDRefine achieves this by employing an adjustable hyperparameter framework, allowing users to optimize trade-offs between experimental data agreement and MD ensemble preservation.

• Using benchmarks, the package demonstrates superiority over individual refinement approaches by reducing discrepancies in conformational dynamics predictions across a variety of systems, including nucleic acid simulations.

• MDRefine’s modular structure offers easy customization, allowing researchers to define specific force-field corrections and forward models as per their experimental requirements.

• Equipped with cross-validation tools, MDRefine finds optimal hyperparameter values automatically, streamlining the refinement process across large and complex molecular datasets.

• The package provides a user-friendly setup, complete with documentation and Jupyter notebooks, enhancing accessibility for researchers in computational biology and chemistry.

@BussiGio @TFroehlking @Piompons
💻Code: https://t.co/kyxK4sSYvU
📜Paper: https://t.co/tgedBBjqTo

#MolecularDynamics #MDRefine #ComputationalBiology #Bioinformatics #ForceField #Python #OpenSource

[Area Science Park @AreaSciencePark]

Can we predict, at very early stages of a viral outbreak, potential mutations in viral proteins, which might emerge in viral variants?
Martin Weigt’s speech opens the second session of the day at #prpconference2024

[Area Science Park @AreaSciencePark]

New dataset (DPCfam-UHGP50) developed by LADE - Data Engineering Laboratory to improve protein sequence annotation and promote discoveries in metagenomics: a valuable resource for a better understanding of the human gastrointestinal proteome.

More info: https://t.co/1p1ov0AII4

[Area Science Park @AreaSciencePark]

Il nuovo studio del laboratorio di Data Engineering utilizza l'#AI per prevedere l'impatto delle mutazioni genetiche sulla stabilità delle proteine, aprendo la strada a innovazioni nella ricerca biomedica >> https://t.co/M60OEDWIi0

[The Journal of Physical Chemistry @JPhysChem]

.@piompons, Miroslav Krepl, @sponer_lab, & @BussiGio investigate the role of #m6A in #RNA recognition! #metadynamics, replica exchange, and alchemical calculations combined to sample RNA:YTH complex conformations, focusing on hydration dynamics.
https://t.co/mcFXRbIBug

[bussilab @bussilab]

📢 Our work on #m6A role in #RNA recognition has been published on @JPhysChem B! https://t.co/LVPU4zyIKh An effort led by @Piompons with help from Miroslav Krepl, @sponer_lab, and @BussiGio Summary in the cited post 👇

[Valerio Piomponi @Piompons]

@gcaravagna @Sissaschool @UniTrieste se non sbaglio quel bagnino è di Piombino

[Potestio Lab @r_potestio]

It was a great pleasure to deliver a LADE seminar at @AreaSciencePark in #Trieste! Big thanks to Francesca Cuturello and Alessio Ansuini @ansuin for having me here!

https://t.co/lNVCevTzQx

[Area Science Park @AreaSciencePark]

#prpconference2024 on pandemic preparedness: meet the speakers!
We are pleased to introduce @Tuliodna who will give a talk on "Using genomics to understand climate amplified diseases and epidemics”.

To learn more on the International Conference: https://t.co/z4TvxDVSqw

[bussilab @bussilab]

Excited to share our new #preprint on #m6A's role in #RNA recognition! 🧬 A collaborative effort by @piompons, Miroslav Krepl, @sponer_lab, & @BussiGio. Dive into our findings in this thread 👇

[Valerio Piomponi @Piompons]

@digitalphotobuz @AOC E weeeeewds eddeededeedeee eefe

[Alessandra Magistrato @Magistrato_Lab]

Postdoc available in my lab starting Dec 2023 on classical and QM/MM MD simulations to refine structures and assign metal ions within cryo-EM maps of RNA molecules.
For info or to apply send me a cover letter & CV!
Project in collaboration with @bussilab and experimentalists.

[bussilab @bussilab]

3 days left to apply for a PhD position in Physics and Chemistry of Biological Systems @SbpSissa @Sissaschool . Application and entrance exam are online. More info 👇Please RT, and apply!

[Valerio Piomponi @Piompons]

Glad to had the opportunity to present part of my PhD work at the"Biophysics of RNA and ribosome" session of #EBSA2023.
@bussilab

[bussilab @bussilab]

Our work presented this week at #EBSA2023 session ‘Biophysics of #RNA and ribosomes’ on Wed morning at 8:30, room B. @Piompons will show how we use #MD, #NMR, and #SAXS to reconstruct the dynamics of A-to-I edited dsRNA. Collab with @Sattler_lab

[bussilab @bussilab]

Our work presented today at the Chemistry department https://t.co/8cINiBqd6U @unitorvergata! @BussiGio will show #RNA dynamics from integrating #MD simulations, #SAXS, #NMR, and #SHAPE, as done by @mattia_bee @Piompons and Nicola Calonaci. Thanks @pcalligari for the invitation!