Yogesh Kalakoti

Ph.D.
Trying to understand proteins

[Yogesh Kalakoti @iam_ysk]

Just in. AFsample2 uses MSA column masking to model alternate protein conformations. Also, you can run most existing methods (just a --flag away) with our pipeline, making it a one stop solution.

Code :
https://t.co/2SAF9Sc3gW

Article: https://t.co/jaUeSsfHTs

[Biology+AI Daily @BiologyAIDaily]

Improving AlphaFold2 Performance in Virtual Screens Targeting GPCRs by Enhancing Binding-Site Conformational Sampling

1. The paper introduces AFsample2T, a targeted AlphaFold2 sampling strategy that boosts virtual screening performance for GPCRs by generating diverse binding-site conformations rather than relying on a single “best” structure.

2. Core idea: mask selected MSA columns only in the orthosteric-site region (extracellular-facing TM segments + part of EL2) to weaken local coevolution constraints and encourage alternative pocket geometries, while keeping the rest of the receptor well-constrained.

3. AFsample2T contrasts with global masking (AFsample2), which reduced binding-site accuracy for GPCRs here (AUC 0.38). Targeted masking preserved fold quality while improving pocket modeling, showing that “where to perturb” matters as much as “how much to perturb”.

4. The authors benchmarked 10 class A GPCRs using 61 curated experimental binding-site structures (from an initial pool of 119 PDB structures) and generated 1,000 models per receptor to quantify how often predicted pockets match experimental ones within 1–2 Å side-chain RMSD (symmetry-aware).

5. Moderate targeted masking improved binding-site accuracy: default AF2 AUC 0.54; adding dropout without masking AUC 0.57; targeted masking at 10–30% reached AUC ~0.59–0.61. Too much masking (50%) degraded secondary structure and collapsed performance (AUC 0.43).

6. The best-performing ensemble (AFsample2T) mixes 250 models each from 0%, 10%, 20%, and 30% masking (with dropout), yielding AUC 0.63 and capturing 73.8% of experimental binding sites at 1.5 Å RMSD vs 60.7% for AF2 (a 22% relative gain at that threshold).

7. A key mechanistic improvement is realistic binding-site plasticity. Median binding-site side-chain RMSF increased from 0.15 Å (AF2; overly rigid pockets) to 0.45 Å (AFsample2T), approaching experimental variability (median 0.58 Å). Backbone RMSF similarly moved from 0.10 Å (AF2) to 0.28 Å (AFsample2T), close to experiment (0.30 Å).

8. AFsample2T also mitigates a known docking issue: AF2 often predicts narrow/collapsed pockets. Across receptors, mean pocket volume increased (209 → 218 Å3) and the “top 1% most open” pockets expanded substantially (272 → 389 Å3), closer to experimental pockets (mean 256 Å3). This was especially relevant for μ-opioid receptor, where AF2 pockets were too collapsed.

9. Virtual screening evaluation used DOCK3.8 with ChEMBL actives (52–202 per receptor) and property-matched ZINC20 decoys, totaling extremely large-scale docking (reported as >240 trillion complexes scored). Rigid-receptor docking was used, making pocket microstates critical.

10. Ensemble screening + ligand-guided model selection is the practical win: while median enrichment of AF2-based models remained below experimental structures, the top 1% AFsample2T models improved early enrichment (mean aLogAUC top 1%: 10.8 → 12.9; mean EF1% top 1%: 7.5 → 9.6). In some targets (e.g., TAAR1, μ-opioid receptor), best AFsample2T models approached top experimental-structure performance.

11. The paper provides a workflow for prospective use: generate ≥250 AFsample2T models for the relevant receptor state, dock a ligand/decoy control set to compute enrichment, select ~top 1% models, manually inspect key interactions/poses, then proceed to large-library prospective screening.

12. Modeling receptor state is handled explicitly: inactive models use receptor sequence alone; active models are generated by cofolding receptor with heterotrimeric G protein sequences via AF2-Multimer, capturing hallmark TM6 movements and separating “state sampling” from “pocket microstate sampling”.

💻Code: https://t.co/DMW991M9ey
📜Paper: https://t.co/jKVzIdt6bx
#AlphaFold2 #GPCR #VirtualScreening #Docking #StructureBasedDrugDesign #ComputationalChemistry #Bioinformatics #ProteinStructure #DrugDiscovery #MachineLearning

[NCIB Headquarters @NCIBHQ]

“दुकानदार ने कहा कि –
बिका हुआ सामान न तो बदला जाएगा और न ही वापस होगा?"

ग़ुस्से में आकर झगड़ा न करें....
क्योंकि आपके पास है 7 दिन का रिटर्न और रिप्लेसमेंट का कानूनी अधिकार।

• चाहे ऑनलाइन शॉपिंग हो या ऑफलाइन स्टोर – सामान ख़राब मिला तो या तो बदलना पड़ेगा या फिर वापस होगा।

• उपभोक्ता संरक्षण अधिनियम, 2019 आपको देता है इसकी पूरी ताक़त। अगर आपका हक़ छीना जाए, तो चुप मत रहिए, राष्ट्रीय उपभोक्ता हेल्पलाइन 1800-11-4000 पर कॉल करके अपनी शिकायत दर्ज कराए।

सादर: @NCIBHQ

[Yogesh Kalakoti @iam_ysk]

@iiserkol @EduMinOfIndia @PTI_News @PIB_India @skkhare781 @priyank_june

[Ben Murrell @BenjMurrell]

My lab, at Karolinska in Stockholm, is looking for a PhD student with a computational/quantitative background to work on probabilistic/generative models of protein (structure and sequence). The research will involve methods development and applications in vaccine design.

[Communications Biology @CommsBio]

AFsample2 predicts multiple conformations and ensembles with AlphaFold2: Communications Biology, Published online: 05 March 2025; doi:10.1038/s42003-025-07791-9AFsample2 improves sampling in AlphaFold, offering structural insights into proteins with… https://t.co/ICgD55zHzS

[ISRO @isro]

Spadex undocking captured from both SDX-1 & SDX-2! 🛰️🛰️🎥

Watch the spectacular views of this successful separation in orbit.

Congratulations to India on this milestone! 🇮🇳✨

#Spadex #ISRO #SpaceTech

[Yogesh Kalakoti @iam_ysk]

@DdelAlamo @RolandDunbrack We ultimately did this. And the results are interesting. https://t.co/wuwfJFz8DQ… PS: MSAsubsampling, if implemented as described in the orig. article, is still the best among older methods.

[Yogesh Kalakoti @iam_ysk]

Just in!! A mini-review summarizing recent progress in predictive models for structural variation.
https://t.co/SksLFsW8IU

[Patrick Bryant @Patrick18287926]

I'm hiring a new postdoc in protein design (wet lab) at Stockholm University/SciLifeLab. The work will focus on expanding our work on binder design and using new AI models that we are developing. Apply here: https://t.co/Y52eedbB0i

[Yogesh Kalakoti @iam_ysk]

@rishibagree natural selection in effect

[Sidhant Sibal @sidhant]

Sweden PM Ulf Kristersson wishes on Diwali 🪔

[Joanna Slusky 🫖 @SluskyLab]

These laureates are VERY deserving. But an important addendum—no model could be trained without 40 years of Helen Berman and others’ work at the PDB, and we would not be certain of AF’s success without 30 years of John Moult and others’ work on CASP. Hats off to them as well 🥂

[Björn Wallner @bjornwallner]

If you want to know how we improved upon today's Nobel Prize in Chemistry. It is out in @NatureComms
https://t.co/zrvd7kyb9U
#NobelPrize2024

[Roland Dunbrack 🏳️‍🌈 @rolanddunbrack.bsky.social @RolandDunbrack]

Why Alphafold succeeded - the best large, curated, cleaned training data set is the PDB. Thank Helen Berman for that.

[International Chess Federation @FIDE_chess]

🇮🇳 India beats 🇨🇳 China in Round 10 of the Women's section at the #ChessOlympiad!
Final score: India 2½ - 1½ China

📷 Michał Walusza

[Yogesh Kalakoti @iam_ysk]

@DdelAlamo @RolandDunbrack Did some initial testing with speach_AF, but it wasn't included in the final list (very long processing time with the provided script, did not go well with sampling). Should revisit though.

In short, no, we've directly compared to only MSAsubsamping, MSAclustering and AFsample.

[Yogesh Kalakoti @iam_ysk]

@POINTNEMkamal As the post mentions, she heads Reliance foundations, the offical sponsor for team India this year and she has been a member of the IOC since 2016. While one can indefinitely argue on the ways to these posts, but the hard fact is that she is the admin face for 🇮🇳 this olympics.

[Yogesh Kalakoti @iam_ysk]

@sokrypton This is similar to what we see with AFsample2, wherein we randomise x% of MSA columns to increase conformational diversity (open/close states). Improvements were observed till around 40% randomization. However, the first row (aa seq) was unchanged.
https://t.co/GrEWtpbPT8