Ep 14: [Journal] WRAPS, MaSIF-designed CID, SHAPES, Evo, EVOLVEpro

今回は、podcast 初企画、岡本、本田、松林でフラッシュ論文紹介を行いました。 本田紹介論文1: Transforming an ATP-dependent enzyme into a dissipative, self-assembling system https://www.nature.com/articles/s41589-024-01811-1 Dr. F. Akif Tezcan: http://tezcan.ucsd.edu/index.php/copy/ Dr. Tuomas Knowles: https://www-knowles.ch.cam.ac.uk/directory/tpjk 小分子のself-assembly system: https://doi.org/10.1016/j.chempr.2019.11.008 松林紹介論文1: Solubilization of Membrane Proteins using designed protein WRAPS https://www.biorxiv.org/content/10.1101/2025.02.04.636539v1 ApoA-Iによる膜タンパク質可溶化(Mizrachi et al., 2015 Nat Commun) https://www.nature.com/articles/ncomms7826 RFDiffusion https://www.nature.com/articles/s41586-023-06415-8 https://github.com/RosettaCommons/RFdiffusion SolubleMPNN https://www.nature.com/articles/s41586-024-07601-y Colab版RF-DiffusionでSolubleMPNNを使う場合は、use_solubleMPNN: にチェック https://colab.research.google.com/github/sokrypton/ColabDesign/blob/main/rf/examples/diffusion.ipynb OmpAなどバクテリアのタンパク質輸送に関する総説 https://www.annualreviews.org/content/journals/10.1146/annurev.biochem.77.061606.160747 Nanodiscs: https://pubs.acs.org/doi/10.1021/nl025623k 岡本紹介論文1 : Targeting protein–ligand neosurfaces with a generalizable deep learning tool https://doi.org/10.1038/s41586-024-08435-4 Correia Lab: https://www.epfl.ch/labs/lpdi/ MaSIF論文(本論文の先行論文): Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning (https://doi.org/10.1038/s41592-019-0666-6 ) 本田紹介論文2: Assessing Generative Model Coverage of Protein Structures with SHAPES https://www.biorxiv.org/content/10.1101/2025.01.09.632260v2 Dr. Possu Huang: http://www.proteindesign.org/people.html TM-score: https://en.wikipedia.org/wiki/Template_modeling_score Protpardelle: https://www.pnas.org/doi/10.1073/pnas.2311500121 Pappardelle (リボン状のパスタ): https://en.wikipedia.org/wiki/Pappardelle MultiFlow: https://github.com/jasonkyuyim/multiflow One-hot encoding: https://en.wikipedia.org/wiki/One-hot TokenizationとEmbedding: 自然言語処理の文脈において、Tokenizationは入力としての文章を扱いやすいユニットに分割すること、数字にEmbeddingはTokenizeされた文章を(Neural networkに入力できるように)高次元のベクトルに変換すること https://geoffrey-geofe.medium.com/tokenization-vs-embedding-understanding-the-differences-and-their-importance-in-nlp-b62718b5964a ESM3: https://github.com/evolutionaryscale/esm FoldSeek: https://github.com/steineggerlab/foldseek Chroma: https://github.com/generatebio/chroma Genie2: https://github.com/aqlaboratory/genie2 CATH: https://www.cathdb.info/ 古賀信康先生: https://researchmap.jp/nykoga All-alpha protein design: https://www.nature.com/articles/s41594-023-01147-9 Nevel alpha-beta protein design: https://www.nature.com/articles/s41594-023-01029-0 「ロボット実験工場」のアイデア (Robotic crowd biology): https://www.nature.com/articles/nbt.3758 松林紹介論文２：Sequence modeling and design from molecular to genome scale with Evo https://www.science.org/doi/10.1126/science.ado9336 https://github.com/evo-design/evo https://evodesign.org/syngenome/about https://aasj.jp/news/watch/25610 収録後に発表された Evo-2 についての preprint と Nature での紹介記事 https://www.biorxiv.org/content/10.1101/2025.02.18.638918v1 Biggest-ever AI biology model writes DNA on demand https://www.nature.com/articles/d41586-025-00531-3 Evo Designer (powered by Evo-2) https://arcinstitute.org/tools/evo/evo-designer StripedHyena https://github.com/togethercomputer/stripedhyena?tab=readme-ov-file pLDDT: LDDTの予測値として⁠AlphaFold2⁠開発過程で開発された指標。値が高いほど予測のconfidenceが高いとみなせる ⁠https://alphafold.ebi.ac.uk/faq 合成ゲノム細菌 JCVI Syn3.0 https://www.science.org/doi/10.1126/science.aad6253 https://www.jcvi.org/research/first-minimal-synthetic-bacterial-cell 岡本紹介論文2 : Rapid in silico directed evolution by a protein language model with EVOLVEpro https://www.science.org/doi/10.1126/science.adr6006 Github: https://github.com/mat10d/EvolvePro Google colab: https://colab.research.google.com/drive/1YCWvR73ItSsJn3P89yk_GY1g5GEJUlgy?usp=sharing Arnold labからのactive learning-assisted directed evolution論文: https://doi.org/10.1038/s41467-025-55987-8