eyeshield21

[arc. @arceyul]

el ingeniero que construyó Claude Code acaba de publicar un video de 28 minutos sobre cómo escribir prompts que realmente funcionan

he visto cursos de 300$ que no cubren lo que él muestra en los primeros 10 minutos

archivos CLAUDE.md, atajos de memoria, sesiones paralelas, patrones de prompting

todo en un video y completamente gratis

funciona seas desarrollador, principiante o alguien que lleva meses usando Claude

[Benetton @benenen12345]

クライオ電顕データセット解析の自動化

CryoSPARC上での生データの撮影像から高分解能の密度マップ生成までの過程を自動化
対象としたGPCRのデータセットのうち8割で分解能が向上
本論文でのパイプラインはダウンロードして利用可能とのこと #BNTNJC
https://t.co/cfXEx9uHmo

[Joruno @wsl8297]

做数据分析或训练模型，最难的往往不是写代码，而是把对的数据找出来。这一步既关键，又最耗时间。

今天在 GitHub 看到一份全球高质量公开数据集清单：Awesome Public Datasets，已拿下 74 k+ Star。

覆盖农业、生物、经济、气候、教育、医疗等数十个领域；每个数据集都标明可用状态，并附元数据链接，找起来更省心。

GitHub：https://t.co/fBfyCg0opd
社区：https://t.co/LA2364Yf5I

更实用的是，它还整理了数据格式、体量、更新频率等关键信息，支持 CSV、JSON、数据库等多种类型。

项目持续维护更新，数据源权威可靠，建议收藏，随用随取。

[SONIA @S0N_IA_]

Este chico te enseña la manera de aprender cualquier cosa 10 veces más rápido con IA

NotebookLM + Gemini + Obsidiana

🔖 guárdatelo lo agradecerás…

[Shunzhi Wang @wang_shunzhi]

1/ Excited to share our new Review in @Nature:
“The past, present and future of de novo protein design.”
Here, we mainly focused on structure-guided protein design. The field is entering a new phase: now that we can design new proteins, what should we build next?

[Eric Topol @EricTopol]

Predicting RNA 3D structure is a grand challenge of biology.
2 new approaches published today to get at it
https://t.co/KQ2oCGlUrN @MolecularCell
https://t.co/ofupia6du2 @NatMachIntell

[Biology+AI Daily @BiologyAIDaily]

Small-molecule binding and sensing with a designed protein family @NatureComms

🚀 New paper from David Baker!🚀

1 Researchers have developed a powerful new computational strategy that combines deep learning with physics-based methods to design a family of proteins capable of binding diverse small molecules with high affinity and atomic-level precision.

2 The core innovation lies in the diversification of the NTF2-like fold using structure generation algorithms to create over 10,000 unique scaffolds with varied internal pocket geometries tailored for ligand docking.

3 This approach successfully produced functional binders for six chemically distinct targets, including hormones like cortisol and drugs such as the anticoagulant apixaban, with binding affinities reaching the nanomolar range.

4 The design process utilized LigandMPNN, a deep learning model trained specifically on protein-ligand complexes, which proved highly effective at generating accurate protein-ligand interactions that were later confirmed by crystal structures.

5 Beyond simple binding, the study demonstrates a modular platform for sensor development by creating a cortisol-induced heterodimerization system that functions as a bioluminescent biosensor at physiologically relevant concentrations.

6 High-resolution structural analysis revealed that the designed proteins match their computational models with sub-angstrom accuracy, validating the reliability of these new deep learning-integrated design pipelines.

7 This work establishes a versatile foundation for designing custom sensors and binders for a wide array of applications across environmental monitoring, diagnostics, and therapeutics.

📜Paper: https://t.co/xCmW8yDkzs
#ProteinDesign #DeepLearning #Biosensors #ComputationalBiology #Biochemistry #NatureCommunications

[Nature Reviews Drug Discovery @NatRevDrugDisc]

The therapeutic potential of orphan adhesion G-protein-coupled receptors
https://t.co/S3VGtldvWY
https://t.co/DYmMOQTthf

This Review discusses opportunities to target aGPCRs in cancer, psychiatric disorders and autoimmune diseases, based on advances in structural biology

[NatureStructMolBiol @NatureSMB]

New online! Structural insight into IscB’s RNA-lid-based inactivation mechanism https://t.co/pIXb8d8OaA

[Biology+AI Daily @BiologyAIDaily]

Co-folding of Membrane Proteins and Lipid Molecules Improves Membrane-Protein Structure Prediction Accuracy

1 This study introduces CoMPLip, a straightforward protocol that feeds lipid molecules into AlphaFold 3 as co‑folded inputs, allowing the model to predict a protein and surrounding lipids simultaneously.

2 During prediction, the lipids self‑assemble into bilayer‑like arrangements around transmembrane helices, thereby providing an explicit membrane environment that was previously only implicitly encoded in training data.

3 In a detailed case study of the E. coli protease RseP bound to batimastat, adding 100 molecules of 1‑monoolein reduced the ligand RMSD from 10.79 Å (no lipids) to 1.37 Å, while the protein backbone remained similarly accurate.

4 A benchmark of 65 experimentally determined membrane‑protein–ligand complexes showed that CoMPLip produced correct ligand poses for four additional targets, raising the number of correct predictions from 41 to 45, though the overall success rate remained unchanged.

5 For 123 full‑length single‑pass proteins, CoMPLip enabled clear extracellular–intracellular domain separation in 61 cases versus only 20 without lipids, demonstrating that the lipid layer physically blocks domain contacts across the membrane.

6 When applied to the dynamic transporter NTCP, CoMPLip sampled both inward‑facing and outward‑facing conformations (approximately 60 % each), whereas standard AlphaFold 3 predicted only the inward state, illustrating enhanced conformational diversity.

7 To mitigate the influence of the many added lipids on the ranking score, the authors devised SCoMPLip, a protein‑ligand‑specific metric that weights protein and ligand pTM scores and the protein–ligand interface, improving model selection for the target complex.

8 Practical guidelines suggest using about 100 copies of 1‑monoolein for a four‑pass transmembrane protein; however, this increases GPU memory demands and may limit applicability to very large targets.

9 The authors note that CoMPLip’s benefits are modest for ligand pose accuracy overall, and that lipid placement can sometimes be unphysical (e.g., occupying the pore); future work will explore other co‑input molecules and broader ranking‑score optimization.

10 Overall, CoMPLip offers a training‑free, easily implementable enhancement for AlphaFold‑based membrane protein modeling, adding an explicit membrane context that improves ligand binding, domain assembly, and state sampling in realistic environments.

📜Paper: https://t.co/N2ZmsYkln4
#ProteinStructure #AlphaFold3 #MembraneProteins #ComputationalBiology #Bioinformatics #StructuralBioinformatics

[SciTech Era @SciTechera]

Quantum computing just touched real biology.

For the first time, scientists at IBM and Cleveland Clinic simulated the electronic structure of a real protein 👀

They used a hybrid system called quantum centric supercomputing (QCSC) to model the 303-atom Trp-cage miniprotein.

Instead of forcing a quantum computer to do everything, they split the problem intelligently:

Classical computers handled the easier regions, while quantum processors tackled the hardest part, strong electron correlations.

Using wavefunction-based embedding (EWF), the protein was broken into smaller quantum-relevant clusters and solved efficiently.

And NOTE, this is NOT protein folding like AlphaFold.

This goes deeper.

It simulates how electrons behave inside molecules, the level that actually determines chemical reactions, drug binding and molecular stability.

Right now, drug discovery still relies on approximations and trial and error.

But if we can simulate molecules at the true quantum level, we move toward designing drugs before lab testing, predicting reactions with high accuracy and even creating entirely new materials.

[Machine Learning in Chemistry @ML_Chem]

Ligand Conformational Variability Enhances Machine Learning Prediction of Protein–Ligand Binding Affinity #machinelearning #compchem https://t.co/DRR0Y4C0Zf

[Matt Dancho (Business Science) @mdancho84]

The 10 types of clustering that all data scientists need to know.

Let's dive in:

[Biology+AI Daily @BiologyAIDaily]

Large Scale Prospective Evaluation of Co‑Folding Across 557 Mac1‑Ligand Complexes and Three Virtual Screens

1. The study prospectively tests AlphaFold3, Chai‑1, and Boltz‑2 on 557 newly crystallized SARS‑CoV‑2 Mac1 inhibitors, achieving >50 % of ligand poses within 2 Å RMSD—outperforming the classic DOCK3.7 benchmark.

2. Pose accuracy remains high even when the test ligands share little structural similarity (low ECFP4 Tanimoto or MCS%) with any ligand in the models’ training sets, indicating genuine generalization rather than memorization.

3. Co‑folding confidence scores (AF3 L‑pLDDT, Chai‑1 ipTM) discriminate accurate poses with AUCs of 0.76 and 0.73, while Boltz‑2’s predicted pIC₅₀ correlates strongly with experimental potency (r = 0.60) and yields MAEs ≈0.7 kcal after linear calibration.

4. When applied to large‑library hit lists (AmpC β‑lactamase, σ₂ receptor, dopamine D4), co‑folding did not surpass docking in separating true binders from false positives; DOCK3.7 maintained higher enrichment and AUCs.

5. The complementary strengths suggest a workflow: use docking for rapid hit identification, then apply co‑folding for accurate pose prediction and affinity ranking within a discovered series to guide optimization.

6. This is the first extensive prospective assessment of co‑folding on a diverse, industrial‑scale dataset, demonstrating that deep‑learning models can reliably inform structure‑based drug design beyond traditional docking.

💻Code: https://t.co/DQK0hnPXI0
📜Paper: https://t.co/fzH9yWYpzp
#DeepLearning #ProteinLigand #StructureBasedDrugDiscovery #CoFolding #AlphaFold3 #Boltz2 #Chai1

[JCIM & JCTC Journals @JCIM_JCTC]

Discovery of a Covalent Small-Molecule eEF1A1 Inhibitor via Structure-Based Virtual Screening #VirtualScreening #DrugDiscovery
https://t.co/K7AHTM2AMO
#JCIM Vol66 Issue2 #compchem

[INOUE lab. (GSFS/ISSP Univ. Tokyo) @LabInoue]

Our paper on inward H+ pumping rhodopsin, Schizorhodopsin (SzR), from Antarctic bacteria—by former member Dr. Marín, Prof. Béjà and Dr. Rozenberg (Technion)—is now published in Biophys. J.!🎊🦠SzRs are conserved across a broad range of species 📄
https://t.co/x262HBOzIr
@BiophysJ

[Stephane Redon @StephaneRedon]

In this thread, explore how to effortlessly prepare coarse-grained systems using GROMACS Wizard. Perfect for molecular modelers!

[Nanome @nanome_inc]

See every chemical interaction in XR
Nanome v2.4 introduces native chemical interaction calculations.
H-bonds. Salt bridges. π-stacking. Halogen bonds. Hydrophobic. Metal coordination. VDW clashes.
Move your ligand and watch interactions update instantly.

[Oz @oznova_]

This is what amino acids look like if you let the side chains flop around on the page

[rcsb pdb 💉🧬💻🔬💊🌱🧠🦠 @buildmodels]

Surprisingly, histones have also been characterized in bacteria, including a bacterium found in soil and aquatic environments and in a spiral-shaped gram-negative bacterium
More at Molecule of the Month: https://t.co/AuVysaeXWw