Professor of Theoretical Chemistry at Sorbonne Universitรฉ & Director @ LCT (UMR 7616 @CNRS)| Co-founder & CSO @qubit_pharma | dev @TINKERtoolsMD (My Views)
Paris, France
Joined March 2012
- Tweets8K
- Following1.8K
- Followers7.2K
- Likes4.6K
Pinned Tweet
๐ New #compchem Preprint ๐
We are excited to share our latest paper: "LAO+: Focused, Multi-Replica Accelerated Orthogonal Sampling for Robust Binding Free Energy Predictions"
arxiv.org/abs/2610.20436
In this work, we introduce LAO+, a multi-replica framework designed to overcome sampling bottlenecks in absolute binding free energy (ABFE) calculations. It uses the polarizable AMOEBA force field and the GPU-accelerated Tinker-HP MD engine. Here are the key highlights:
โก 36x Faster Convergence: By using a partitioned lambda-dynamics scheme and Hamiltonian replica exchange (HREX), LAO+ achieves a 4-fold speedup over our previous LAO method, and up to ~36-fold over standard Lambda-ABF.
๐ฏ High Accuracy: Evaluated across 28 protein-ligand complexes (including BRD4, Trypsin, CHK1, PWWP1, and P38), LAO+ achieves an RMSE of 0.83 kcal/mol and a 0.87 correlation with experiment.
โฑ๏ธ Ultra-Efficient: High thermodynamic precision is obtained from just 63 ns of total sampling per ligand. ๐ Conformational Reservoirs: Specialized biased reservoirs easily cross high orthogonal barriers, recovering hidden protein side-chain and ligand rotational states (e.g. T4 lysozyme, BRD4, and JNK1).
By resolving slow hydration and conformational traps, LAO+ makes rigorous ABFE calculations truly practical for computer-aided drug design. Congratulations to all co-authors: @Narjes_Ansari, Florent Hรฉdin, Jรฉrรดme Hรฉnin, and Louis Lagardรจre. @qubit_pharma
#compchem Good read: Why Interfacial Electric Fields May Appear Weaker: An Electron-Density-Based Study of the Flat AirโWater Interface.
pubs.acs.org/jctcce/article/โฆ
#compchem Good read: Physics-based molecular fingerprints from spectral graph theory provide efficient geometry-aware measures of chemical similarity.
pubs.rsc.org/sc/article/doi/โฆ
Good read: The Navier-Stokes equations blow up, and planes still fly.
Assuming the Navier-Stokes problem is solved, what actually changes for physics, and what stays the same?
A @CommsPhys Editorial.
#NavierStokes #FluidDynamics #AI
nature.com/articles/s42005-0โฆ
Jean-Philip Piquemal retweeted
[HOMMAGE] Le CNRS a appris avec une grande tristesse le dรฉcรจs de Claude Cohen-Tannoudji. Grande figure de la physique atomique et de lโoptique quantique mondiales, il a reรงu la mรฉdaille dโor du CNRS en 1996 et le prix Nobel de physique en 1997.
๐ inp.cnrs.fr/fr/cnrsinfo/hommโฆ
Our recent work has been featured for october 2026 by @CommsPhys #quantumcomputing #compchem
nature.com/commsphys/
New #quantumcomputing paper published in Communications Physics (Nature Portfolio):
"๐๐ง ๐จ๐ฉ๐ญ๐ข๐ฆ๐ข๐ณ๐๐ ๐๐จ๐ง๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ข๐จ๐ง ๐จ๐ ๐ฅ๐ข๐ ๐๐ฅ๐ ๐๐๐ซ๐ ๐ ๐๐ง๐๐ซ๐๐ญ๐จ๐ซ ๐ฉ๐จ๐จ๐ฅ๐ฌ ๐๐จ๐ซ ๐ฏ๐๐ซ๐ข๐๐ญ๐ข๐จ๐ง๐๐ฅ ๐ช๐ฎ๐๐ง๐ญ๐ฎ๐ฆ ๐๐ข๐ ๐๐ง๐ฌ๐จ๐ฅ๐ฏ๐๐ซ๐ฌ ๐ข๐ง ๐๐ก๐๐ฆ๐ข๐ฌ๐ญ๐ซ๐ฒ".
nature.com/articles/s42005-0โฆ
๐ฌ The evolution of quantum states relies on unitary transformations dictated by Lie algebras. Finding a minimal set of generators to construct these algebras, known as a minimal complete pool (MCP), is a central challenge in the field. Traditionally, the search for these generators has relied on greedy construction steps applied to an exponentially growing number of candidate operators, creating an ๐ข๐ง๐ญ๐ซ๐๐๐ญ๐๐๐ฅ๐ ๐๐จ๐ฆ๐ฉ๐ฎ๐ญ๐๐ญ๐ข๐จ๐ง๐๐ฅ ๐๐จ๐ญ๐ญ๐ฅ๐๐ง๐๐๐ค ๐๐จ๐ซ ๐๐ฏ๐๐ง ๐ฆ๐จ๐๐๐ซ๐๐ญ๐๐ฅ๐ฒ ๐ฌ๐ข๐ณ๐๐ ๐ฌ๐ฒ๐ฌ๐ญ๐๐ฆ๐ฌ.
โกWe propose ๐ ๐ซ๐จ๐๐ฎ๐ฌ๐ญ ๐ง๐๐ฐ ๐ฆ๐๐ญ๐ก๐๐ฆ๐๐ญ๐ข๐๐๐ฅ ๐๐ซ๐๐ฆ๐๐ฐ๐จ๐ซ๐ค ๐๐๐๐ซ๐๐ฌ๐ฌ๐๐ฌ ๐ญ๐ก๐๐ฌ๐ ๐๐ก๐๐ฅ๐ฅ๐๐ง๐ ๐๐ฌ ๐ก๐๐๐-๐จ๐ง, ๐๐๐ฅ๐ข๐ฏ๐๐ซ๐ข๐ง๐ ๐ ๐ฉ๐จ๐ฅ๐ฒ๐ง๐จ๐ฆ๐ข๐๐ฅ-๐ฌ๐๐๐ฅ๐ข๐ง๐ ๐ฌ๐ญ๐ซ๐๐ญ๐๐ ๐ฒ ๐ญ๐ก๐๐ญ ๐๐ฒ๐ฉ๐๐ฌ๐ฌ๐๐ฌ ๐ญ๐ก๐ ๐จ๐ฅ๐ ๐๐จ๐ฆ๐ฉ๐ฎ๐ญ๐๐ญ๐ข๐จ๐ง๐๐ฅ ๐ฐ๐๐ฅ๐ฅ๐ฌ. As an application, we show that these MCPs allow for improved Variational Quantum Eigensolver (VQE) strategies for quantum chemistry.
Beyond that, this foundational framework also presents broad applications across quantum computing, including quantum error correction, machine learning, and hardware control. ๐
Great work by Y. Viswanathan, Olivier ADJOUA, Cรฉsar Feniou and Siwar Badreddine, PhD.
Sorbonne Universitรฉ / CNRS
@qubit_pharma
#compchem #liealgebra #vqe
#AI Good read: Three reasons why AI is not conscious and perhaps never will be.
nature.com/articles/d41586-0โฆ
We have an opening in my group at @LCT_UMR7616 & @qubit_pharma for a postdoc in #machinelearning within the context of the FeNNix-Bio1 & large scale reactive molecular simulations.
You can have an idea of the type of methodology we use by checking the preprint here.
chemrxiv.org/doi/abs/10.2643โฆ
Contact me for more details.
#compchem #FeNNixBio1
Jean-Philip Piquemal retweeted
LAO+: Focused, Multi-Replica Accelerated Orthogonal Sampling for Robust Binding Free Energy Predictions
Narjes Ansari, Florent Hรฉdin, Jรฉrรดme Hรฉnin, Jean-Philip Piquemal, Louis Lagardรจre
arxiv.org/abs/2610.20436 [๐๐๐ข๐๐๐๐.๐๐๐๐-๐๐]
ALT We introduce Lambda-ABF-OPES+ (LAO+), a multi-replica framework for absolute binding free energy (ABFE) calculations that accelerates both the alchemical transformation and the slow โorthogonalโ motions coupled to it. At the core of LAO+ is a partitioned ฮป-dynamics scheme that concentrates sampling on the abrupt rehydration of the binding pocket, the main bottleneck of ligand decoupling, while Hamiltonian replica exchange relays configurations along the entire alchemical path. This drives a fourfold gain in convergence speed for the rate-limiting leg over our previous LAO method, which was itself up to ninefold faster than Lambda-ABF, an overall gain of up to \sim36-fold. Applied to 28 proteinโligand complexes across five targets, LAO+ achieves a root-mean-square error of 0.83 kcal mol^-1 and a correlation of 0.87 with experiment from only 63 ns of sampling per ligand. Moreover, dedicated reservoir replicas accelerate any slow degree of freedom that can be described by a collective var
Jean-Philip Piquemal retweeted
๐๐๐ LAO+ / Tinker-HP / @nvidia GPUs : X 36 acceleration compared to Lambda-ABF.
๐ New #compchem Preprint ๐
We are excited to share our latest paper: "LAO+: Focused, Multi-Replica Accelerated Orthogonal Sampling for Robust Binding Free Energy Predictions"
arxiv.org/abs/2610.20436
In this work, we introduce LAO+, a multi-replica framework designed to overcome sampling bottlenecks in absolute binding free energy (ABFE) calculations. It uses the polarizable AMOEBA force field and the GPU-accelerated Tinker-HP MD engine. Here are the key highlights:
โก 36x Faster Convergence: By using a partitioned lambda-dynamics scheme and Hamiltonian replica exchange (HREX), LAO+ achieves a 4-fold speedup over our previous LAO method, and up to ~36-fold over standard Lambda-ABF.
๐ฏ High Accuracy: Evaluated across 28 protein-ligand complexes (including BRD4, Trypsin, CHK1, PWWP1, and P38), LAO+ achieves an RMSE of 0.83 kcal/mol and a 0.87 correlation with experiment.
โฑ๏ธ Ultra-Efficient: High thermodynamic precision is obtained from just 63 ns of total sampling per ligand. ๐ Conformational Reservoirs: Specialized biased reservoirs easily cross high orthogonal barriers, recovering hidden protein side-chain and ligand rotational states (e.g. T4 lysozyme, BRD4, and JNK1).
By resolving slow hydration and conformational traps, LAO+ makes rigorous ABFE calculations truly practical for computer-aided drug design. Congratulations to all co-authors: @Narjes_Ansari, Florent Hรฉdin, Jรฉrรดme Hรฉnin, and Louis Lagardรจre. @qubit_pharma
Be part of WATOC 2028!
Pre-register now and join the global theoretical and computational chemistry community.
docs.google.com/forms/d/e/1Fโฆ
#WATOC2028 #TheoreticalChemistry #ComputationalChemistry
#compchem Good read: Evolution of dipoleโdipole dynamics in cold ammonia collisions nature.com/articles/s41557-0โฆ
