Transit at LAX to Santa Fe, for the MLCM-25.

We recently construct a model to evaluate long-range cooperative efforts across protein backbone hydrogen bonds. Here we use it to compare MLFFs including ANI, MACE and Orb, and their performance varies wildly...

We performed individual docking with AlphaFold3 and Vina for 27 DUD-E targets and update the results. It turns out the enrichment factor doubles compared to standard docking.
www.biorxiv.org/content/10.1...

Flew all the way from San Diego back to Hangzhou, and had slept all-day long today. It's nice to meet good scientists and to present in an igloo. Now back to my daily caveman life of writing papers and mentoring students.

Start packing for my trip to SD for #ACSSpring2025 . Will present our JCTC paper of getting most probable transition path (MPTP) with CAR, and looking forward to the nice weather in SD. It's cold these days in Hangzhou.

We developed a transfer learning method that integrates public AMP data and a small number (~300) of activity data of N-terminal modified peptide. The model is quite accurate & it's nice to see how features in latent space converge to known domains. Noise is the key!

www.nature.com/articles/s41...

A detailed look at fresh samples taken from asteroid Bennu confirms that organic chemistry (and the molecules of life) are everywhere:

PIWI proteins undergo large-scale conformational change to recognize and cleave RNA. Congrats to my colleague Prof. En-zhi Shen on this excellent work, now published in Nature! Also very happy that my student Xiaoli Lu and me can contribute.

The new #CECAM program is now live!
76 exciting workshops, schools and conferences will be held across the network between April 2025 and March 2026!
Explore our activities and apply to participate at: www.cecam.org/program

This is just to alert you that SBDD2025 will happen!
Dates are 1-3 October, usual wonderful location, in Sestri Levante - Italy!
sbdd-congress.it

Stay tuned for more news on the program, etc.
Registration will open soon with minimal fees, but also limited sits!

💪🤗

I love doing stupid things. What if, given a compound library, predict the ligand-protein structure for each ligand, then perform docking using these individially generated protein pockets? We did it, and here are the results:

Finding the path connecting two given states is a classical problem. We recently note that imposing hard constraints can be advantageous compared to restraints or reparametrization for path update. Here an example on computing the MPTP in realistic biomolecular systems
doi.org/10.1021/acs....