Kui Xu

Postdoctoral Scholar @ Tsinghua University — Cryo-EM and Structure Biology, 3D Vision and Deep Learning. Banner credit: TL

[Kui Xu @kuixut]

@pushmeet @MirowskiPiotr @GoogleDeepMind @IsomorphicLabs Which type of GPU used on the server?

[Kui Xu @kuixut]

@pushmeet @MirowskiPiotr @GoogleDeepMind @IsomorphicLabs Super cool!!! Super fast!!!

[Kui Xu @kuixut]

Super fast！By using diffusion model with new designed network and the training with 256 A100 GPUs, AlphaFold3 predicts a 5000AA
protein complex within only 20 minutes! #AlphaFold3 #diffusion #deepmind

[Kui Xu @kuixut]

@sokrypton really fast！and memory efficient！

[Kui Xu @kuixut]

Congratulations to Kaiming, Jian Sun, Shaoqing, and Xiangyu, inventors of ResNets, for winning the Future Science Prize. https://t.co/6pUIRxhzKC

[Kui Xu @kuixut]

How much faster when using #mmseq2 in #AlphaFold ?
#ChatGPT says 4 times !!!

It's that right?

[Kui Xu @kuixut]

@sokrypton Any good news about the 214million MSAs?

[Kui Xu @kuixut]

@doubledaibo Amazing！！！

[Kui Xu @kuixut]

@ZhongingAlong @Princeton Congratulations! Ellen, I am looking forward to your next awesome methods on cryo-EM!!!

[Kui Xu @kuixut]

@sukritsingh92 nice findings!!!😆

[Kui Xu @kuixut]

@sokrypton RGN2(https://t.co/ULzcP0t1Ln) is the first to attempt to train a protein language model (AminoBert) for protein structure prediction. @peng_illinois What's the difference between AminoBert and the PLM in OmegaFold? Larger sequence DBs? or pretraining with structure information?

[Kui Xu @kuixut]

Our #CryoNet web server is updated! By introducing homology template modeling, CryoNet could build highly accurate full atomic #protein structure for #cryoEM map from high resolution at 3Å to low at 5~8Å.
Feel free to access CryoNet at: https://t.co/057EMcWVpl.
By CryoNet Team.

[Kui Xu @kuixut]

Adding a big enough number of polyA linker is the same as adding a big enough number for "residue_index".
#AlphaFold #alphafold2

[Minkyung Baek @minkbaek]

Adding a big enough number for "residue_index" feature is enough to model hetero-complex using AlphaFold (green&cyan: crystal structure / magenta: predicted model w/ residue_index modification).
#AlphaFold #alphafold2

[Kui Xu @kuixut]

Our PrismNet paper is online on Cell Research @cell_res! Congratulations to Lei, Wenze, and Yucheng! Thanks to Cliff and all the team members!

Code: https://t.co/kEsalILPAv

https://t.co/0AjDcYxV6u

[Kui Xu @kuixut]

Our #Cell paper is online! Proud to apply our PrismNet to predict proteins binding to #SARS_CoV_2!

PrismNet: https://t.co/kEsalILPAv

In vivo structural characterization of the #SARS_CoV_2 RNA genome identifies host proteins vulnerable to repurposed drugs https://t.co/5eSzEPgogk

[Kui Xu @kuixut]

One paper accepted to Science. Proud to apply our improved A2-Net to identify new proteins in the minor spliceosome, congratulations to Rui Bai and Ruixue Wan.

A2-Net: Building Structure from #CryoEM Density
https://t.co/6BUx3ugQM8

https://t.co/7iEIcBuaYT

[Kui Xu @kuixut]

Our work on the In vivo RNA structure of #SARS_CoV_2 is accepted by Cell today! Proud to apply our deep learning-based method--PrismNet to do the prediction. Congratulations to Lei Sun and Panpan and all other authors!

[Kui Xu @kuixut]

In vivo structural characterization of the whole SARS-CoV-2 RNA genome identifies host cell target proteins vulnerable to re-purposed drugs https://t.co/QUI95MJoL2
@biorxivpreprint

@SARSCoV192

@Sarscov2science

[Kui Xu @kuixut]

What a cool neural interpolation!!!

[AK @_akhaliq]

X-Fields: Implicit Neural View-, Light- and Time-Image Interpolation
pdf: https://t.co/bxOACgBPRz
demo: https://t.co/iAdd6VRbfP
project page: https://t.co/MafCsMmF6K

[Kui Xu @kuixut]

VRmol is online on @OUPBioinfo! Free to explore and perform #drug docking, genomic variation, structure editing, etc. on #3D molecular in #VR.🤟🤟🤟
Many thanks to Nan, Jingle, Xun, and David!!!
https://t.co/0wbMgf7tIA

https://t.co/9CyHVaAqZW

@RtoVR @UploadVR @htcvive @oculus

[Kui Xu @kuixut]

@mozillareality @htc @Viveport #FirefoxReality is also well supporting https://t.co/0wbMgf7tIA, which provide #VR on visualizing molecular structure.