Shiva Shankar S

| Structural Biology and Mass Spectrometry enthusiast | Protein engineering and Proteomics Specialist |
PhD student @csir_ncl

[Shiva Shankar S @BioShankar]

I am happy to share that our work on Metrn and Metrnl structure prediction, domain identification and SNP analysis got recently published in the #jbsd thanks to @KulkarniMj @reemabanarjee and all my co-authors @csir_ncl @AcSIR_India
#meteorin #metrnl
https://t.co/VXZLCGgUNd

[Hira_Labo🐻データサイエンス @hirat62363752]

ORIGAMI: Orientation-Aware Graph Neural Network for Assessing Multimeric Interfaces of Protein Complex Structures

タンパク質複合体予測構造の界面品質をオリエンテーション認識GNNで評価するツール
複合体予測構造の信頼性評価を改善

https://t.co/EBwEhK0MT2
https://t.co/fmbBr38XoW

[Pranam Chatterjee @pranamanam]

What if a binder could not only bind a target, but selectively recognize one conformational state over another? Today, we’re excited to share AlloGen, our experimentally-validated framework for conformation-selective binder design! 🐙

📜: https://t.co/NMq7SZBfAW
🤗: https://t.co/qQ9so68mLY

🧵👇🏾

[Biology+AI Daily @BiologyAIDaily]

HonestAffinity: Leak-Aware Evaluation of Protein and Pocket Priors for Binding Affinity Prediction

1. HonestAffinity frames a key caution for protein–ligand affinity models: architectural “priors” can flip from helpful to harmful depending on whether evaluation splits leak protein/ligand similarity (canonical CASF/PDBbind-style) or are leak-proof (LP-PDBBind 3-tier no-leak).

2. The paper isolates two common priors in a controlled 1D-input setting: frozen ESM-2 (650M) per-residue embeddings (1280-d projected to 256) and a learned binary pocket-position marker added to residue features when pocket annotations are available.

3. Core result is a split-conditioned reversal across both priors. On familiar/canonical splits (val, CASF-2016, CASF-2016 non-train), adding ESM-2 and the pocket marker improves performance; on strict LP no-leak tiers (test_cl1–cl3), the same additions reduce Pearson R.

4. Three deployment-matched variants are proposed rather than one “best” model: HONESTAFFINITY-POCKET (ESM-2 + pocket marker) for familiar/annotated targets; HONESTAFFINITY-NOPOCKET (ESM-2 only) when no pocket list exists; HONESTAFFINITY-POCKET-NOESM (21-token residue embedding + pocket marker) for strict LP-style generalization with pocket annotations.

5. Quantitatively (Pearson R, mean±std over 3 seeds): HONESTAFFINITY-POCKET leads on val (0.548), CASF-2016 (0.747), and CASF non-train (0.646). But on LP strict tiers, HONESTAFFINITY-POCKET-NOESM leads: cl1 0.531, cl2 0.538, cl3 0.497, also giving best RMSE on cl2/cl3.

6. The ESM-2 ablation is especially instructive: swapping ESM-2 for a learned 21-vocab residue embedding decreases R on val/CASF (e.g., CASF-2016 drops from 0.747 to 0.713) but increases R on every strict LP tier (e.g., cl3 rises from 0.433 to 0.497; ∆R up to +0.064).

7. The pocket marker shows the same sign flip: it helps on val/CASF but hurts on LP tiers. Interpretation: both priors inject signals correlated with training-distribution structure (protein-family signatures or pocket geometry), which can become misleading when similarity filtering removes overlap by design.

8. Methodologically, the authors argue for paired reporting: canonical metrics alone would over-credit ESM/pocket priors; leak-proof metrics alone would understate their utility for in-distribution scoring. They recommend routine paired canonical + leak-proof ablations for new affinity predictors.

9. Implementation is intentionally compact and scalable: multi-scale 1D CNNs + a residual block + a single Transformer layer per branch (protein and SMILES), coupled by a matrix-product compatibility map. Training uses 11,513 LP-PDBBind train complexes and runs in ~3 GPU-hours on a single V100; inference is ~10 ms/complex with cached embeddings.

📜Paper: https://t.co/bcmQuo1KRR
#CompBio #Bioinformatics #DrugDiscovery #ProteinLanguageModels #ESM2 #BindingAffinity #Benchmarking #PDBbind #CASF2016 #MachineLearning

[Biology+AI Daily @BiologyAIDaily]

DeepRank-Ab: A scoring function for antibody-antigen complexes based on geometric deep learning

1. DeepRank-Ab targets a key failure mode in antibody–antigen modeling: good poses are often sampled but not ranked correctly. The paper shows an example where AlphaFold3 generates a DockQ 0.6 model yet ranks it 498th, illustrating that scoring (not only sampling) is a central bottleneck.

2. The authors introduce a rigorously curated benchmark tailored to antibody–antigen interfaces: ~2.3 million decoys from 1,442 complexes (from SAbDab), generated under four HADDOCK3 protocols (bound/unbound docking and refinement). Complexes are clustered with Foldseek-Multimer and split at the cluster level to reduce data leakage.

3. DeepRank-Ab is trained as a regression model to predict continuous DockQ (pDockQ), rather than a binary “acceptable/not acceptable” label. This preserves the full spectrum of model quality and directly supports ranking.

4. A major design finding: atom-level interface graphs outperform residue-level graphs for antibody–antigen scoring, consistent with the importance of fine-grained CDR loop geometry. The best-performing formulations combine atom-level nodes with interface-focused edges.

5. The interface representation is strengthened using Voronoi-based surface decomposition: edges can encode either simple interatomic distances or Voronoi contact areas. Across cross-validation, Voronoi contact area provides a consistent edge over distance-only encoding for predicting and classifying near-native poses.

6. Antibody-specific and physics-inspired features matter. Adding IMGT-derived region labels (e.g., CDR annotations) and geometric/energetic descriptors improves performance and stability; ablations show large error increases when removing region labels, atom type, covalent interactions, orientation, or Voronoi contact area. In contrast, raw ESM-2 embeddings are partly redundant, and antibody-specific fine-tuning (AbTune) yields only marginal gains relative to added complexity.

7. Training data sampling is treated as a first-class modeling choice. Besides “balanced” sampling across DockQ bins, they test targeted upsampling of low-quality decoys (DockQ < 0.23) to increase structural diversity in failure modes; this slightly improves overall learning and robustness.

8. Architecture: an E(n)-equivariant GNN with a two-branch design that processes interface edges and internal edges separately, then pools with global attention and predicts pDockQ. This keeps rotational/translational symmetry while learning local interface geometry.

9. On the hardest docking benchmark setting (unbound–unbound docking; n=215), DeepRank-Ab variants outperform HADDOCK EMscoring and VoroIF-jury in Top-k success. Atom-distance edges do best at Top1, while Voronoi-area edges become strongest at higher k (e.g., Top10), reflecting tradeoffs between sharp early ranking and broader enrichment.

10. Generalization beyond docking decoys: on an AF3-focused external test set (59 complexes released after AF3’s training cutoff), DeepRank-Ab improves AF3 Top1 success by 35.5% and more than doubles mean Top1 DockQ. A notable practical tweak is removing the electrostatics term to better handle unrelaxed AF3 structures (distribution shift from clashes/packing), boosting Top1 success to 54.24%.

11. External CAPRI/MassiveFold evaluation (5 antibody–antigen targets) suggests strong out-of-distribution performance: DeepRank-Ab reaches 100% Top5 success, with especially strong results on peptide antigens. The paper also analyzes why: peptide interfaces tend to be more locally driven and compatible with the model’s 5 Å heavy-atom cutoff graph construction.

12. The work argues that “consensus scoring” is not automatically beneficial: combining DeepRank-Ab with AF3 via averaging or jury voting reduces performance, implying DeepRank-Ab already captures much of the useful ranking signal for these systems.

💻Code: https://t.co/qfHjbNkCGs
📜Paper: https://t.co/RWdeEu450N
#ComputationalBiology #StructuralBioinformatics #Antibodies #ProteinDocking #GeometricDeepLearning #GNN #AlphaFold3 #HADDOCK #CAPRI #BenchmarkDatasets

[biohub @biohub]

Designing a protein binder used to mean years of lab experiments. ESMFold2 lets researchers run hundreds of thousands of designs computationally—then take only the most promising into the lab. We tested it across 5 targets in oncology and immunology. It worked.

Download the model and start building: https://t.co/odrOR3U1hj

[Huiluo Cao 曹慧荦 @hlcao]

Treatment of Shiga toxin–producing E. coli infection by CRISPR-Cas–targeted cleavage of the Shiga toxin gene in animal models | Science Translational Medicine https://t.co/elBVvyoktj

[ANRF-National cryo-EM Facility @IITK @GpcrLab3434]

A structure of CXCR4 in complex with a de-novo designed miniprotein blocker determined at our @ANRFIndia facility @IITKanpur is published in @Nature! Spearheaded by the lab of Prof. David Baker (Nobel Prize, 2024) @UWproteindesign, it also involves @arshukla and team @IITKanpur.

[Biology+AI Daily @BiologyAIDaily]

Fine-Tuning Diffusion Models for Molecular Generation via Reinforcement Learning and Fast Sampling

1. FTDiff is a reinforcement-learning fine-tuning framework that aligns a pretrained 3D, target-conditioned diffusion model with structure-based drug design objectives, aiming to generate ligands that dock well while staying drug-like and synthesizable—without expensive post-hoc optimization during sampling.

2. Key idea: treat diffusion denoising as a multi-step policy. The “state” is (protein pocket, timestep, noisy ligand), the “action” is the next denoised ligand, and reward is computed only on the final molecule, enabling direct optimization of design goals rather than only reconstruction-from-noise.

3. Stability and sample efficiency come from a GRPO-style optimization scheme: advantages are normalized within groups of samples per pocket (using group mean/std), avoiding a learned value function and reducing overhead while keeping updates stable under sparse terminal rewards.

4. A practical RL objective for molecular diffusion: the policy factorizes into coordinate and atom-type transitions, and FTDiff uses a Double-Head Clipped Loss (LDHC) with PPO-style clipping applied separately to coordinate likelihood ratios and atom-type likelihood ratios; a “rectified” variant further clips per-coordinate components to improve training behavior.

5. Speed is addressed with a time-free fast sampling mechanism that reduces denoising from ~1000 steps to 10 steps. Instead of relying on explicit timestep schedules (which can break molecular validity when aggressively skipped), FTDiff uses adaptive interpolation coefficients based on denoising “confidence” (coordinate discrepancy and atom-type KL), synchronizing coordinate/type denoising.

6. Multi-objective optimization is driven by a threshold-aware reward: each normalized objective score is passed through a sigmoid centered at an acceptability threshold (e.g., 0.5). This emphasizes improvements near “good-enough” regions and reduces over-optimizing one metric at the expense of others (common in linear weighted sums).

7. Experimental setup: pretrained on CrossDocked2020 (~100k filtered ligand–pocket pairs). Fine-tuning uses 100 representative pockets selected via ESM-2 pocket embeddings + K-means clustering to cover diverse pocket space efficiently; evaluation is on 100 held-out protein targets.

8. Multi-objective results on CrossDocked2020: FTDiff reports Avg Vina Score -7.18, Avg QED 0.56, Avg SA 0.61, and diversity 0.72, while also achieving a high-affinity rate of 78.6% (median 88.9%). Under strict joint constraints (Vina<-8, QED>0.5, SA>0.5), it reaches a 22.3% pass rate with 95.8% connectivity (no broken bonds).

9. Single-objective affinity comparison vs KGDiff (gradient-guided sampling): FTDiff slightly improves docking metrics (e.g., Avg Vina Score -8.11 vs -8.04) and high-affinity rate (89% vs 87%), while keeping similar diversity, suggesting RL fine-tuning can compete with gradient-guided sampling without requiring auxiliary predictors at inference.

📜Paper: https://t.co/y8WAh88BaM
#ComputationalBiology #DrugDiscovery #SBDD #DiffusionModels #ReinforcementLearning #MolecularGeneration #GenerativeAI #Cheminformatics #MachineLearning

[Bowen Jing @bjing2016]

SwitchCraft optimizes sequences under co-folding losses to design proteins that switch states dependent on ligand binding. By composing multiple forward passes and a versatile toolbox of loss functions, we can encode complex input-output relationships. (2/7)

[Thomas Hayes @THayes427]

We’re excited to share the full binder design protocol. Check it out here: https://t.co/AtkipkiYtS.

The notebook includes support for @modal to easily scale up binder generation.

Give it a try and let us know how it works!

You can read more about ESMFold2, ESMC, ESM Atlas, and the full results in the paper here: https://t.co/M3rt00pU8Z.

[Hira_Labo🐻データサイエンス @hirat62363752]

SwitchCraft: A Programmatic Framework for Designing State-Switching Proteins

複数状態が切り替わるタンパク質を設計するためのフレームワーク
任意の標的分子に対する蛍光バイオセンサー設計のためのインシリコ戦略を提供

https://t.co/Qm4yYx6QVH
https://t.co/J5Ka7fnM9q

[Polly Fordyce @fordycelab]

Characterizing AI-designed proteins requires quantitative biochemistry at massive scale. Enter Amplicon/Protein Bead Display (APB-Display), a fully in vitro platform that quantifies Kd's for &gt;100,000 variants in &lt;3 days (preprint link below!) @Stanford_ChEMH @czbiohub (1/n)

[Kresten Lindorff-Larsen @LindorffLarsen]

StruCloze: A Unified Framework for Backmapping and Inpainting Biomolecule Structures

Fast, accurate backmapping from various coarse-grained models for proteins and nucleic acids, including in complexes and condensates

https://t.co/auSk3I4YJH

[Kresten Lindorff-Larsen @LindorffLarsen]

Everything you wanted to know about the protein chemistry behind how amino-acid changes affect the cellular abundance of proteins

Effects of residue substitutions on the cellular abundance of proteins
https://t.co/FwoD1kiF30

[Martin Pacesa @MartinPacesa]

Amazing work from the Kellogg lab!! I had this kind of project in my very first 2022 professorship proposal and but never actually did it, super glad to see it can work in reality!!!

https://t.co/iqofxDnPGP

[Hira_Labo🐻データサイエンス @hirat62363752]

Affinity Fine-Tuning of Boltz-2: An Open Framework for Protein-Ligand Potency Prediction in Drug Discovery

Boltz-2のリガンド親和性予測のファインチューニングフレームワーク
固有の実験データでアフィニティ予測部分のみ最適化

https://t.co/T5htMQi0bP
https://t.co/41SAJMWtJq

[Ayushman Mallick @AyushmanMallick]

Built BoltzYML: an open-source tool to automate Boltz-2 input prep for ternary (Protein + Ligand + Protein) binding prediction
2 CIF files in → ready-to-run YAML out. Sequences, pocket contacts, affinity block: all automated.
🌐 https://t.co/nUtA6ArmBq
Python CLI also
Thread 👇

[Benetton @benenen12345]

自律的な生物発光システムに関する総説

ルシフェラーゼとその基質合成系を導入することで、基質の外部供給なしに発光する系について、研究が進んでいる細菌と真菌型ルシフェラーゼの改変や現状の課題、次世代発光システムなどへの応用の展開などが紹介されている #BNTNJC
https://t.co/dijhEAdxLJ

[Tominaga K. (tomiken) @pacyc1841]

PlasmidGPT: A generative framework for plasmid analysis and generation | Science Advances https://t.co/o8Rzlho73N

[Science Magazine @ScienceMagazine]

Key human defenses against pathogens were forged billions of years ago in microbial battles with viruses.

Learn more: https://t.co/NqwQk38nzX @NewsfromScience