Lightnews — Scholar-powered news

LightNews

lindsey guan

@lindseyguan.bsky.social

680 followers 140 following 10 posts

she/her | PhD student in the Keating Lab @MIT | Bio & CS @UCBerkeley

Posts Media Videos Starter Packs

lindsey guan @lindseyguan.bsky.social · 22h

now out in Protein Science! onlinelibrary.wiley.com/doi/10.1002/...

Training bias and sequence alignments shape protein–peptide docking by AlphaFold and related methods

Protein-peptide interactions mediate many biological processes. Accurate structural models of protein-peptide complexes, determined by experiment or computational prediction, are essential for unders...

onlinelibrary.wiley.com

Reposted by lindsey guan

Nature @nature.com · 20d

We used machine-learning tools in an attempt to recreate the method for cutting funding, and then applied it to past US NIH grants to reveal the broad-reaching consequences

go.nature.com/4pBArFH

What research might be lost after the NIH’s cuts? Nature trained a bot to find out

We used machine-learning tools in an attempt to recreate the method for cutting funding, and then applied it to past US National Institutes of Health grants to reveal the broad-reaching consequences of such action.

Reposted by lindsey guan

Eli Lipsitz @eli.lipsitz.net · Aug 21

The Game Bub Crowd Supply campaign is live! And it's been redesigned with a new horizontal design.

You can now back the campaign to get your very own pre-assembled Game Bub!

www.crowdsupply.com/second-bedro...

An open-source FPGA retro-emulation handheld

www.crowdsupply.com

Reposted by lindsey guan

Yo Akiyama @yoakiyama.bsky.social · Aug 5

Excited to share work with
Zhidian Zhang, @milot.bsky.social, @martinsteinegger.bsky.social, and @sokrypton.org
biorxiv.org/content/10.1...
TLDR: We introduce MSA Pairformer, a 111M parameter protein language model that challenges the scaling paradigm in self-supervised protein language modeling🧵

Scaling down protein language modeling with MSA Pairformer

Recent efforts in protein language modeling have focused on scaling single-sequence models and their training data, requiring vast compute resources that limit accessibility. Although models that use ...

lindsey guan @lindseyguan.bsky.social · Jun 24

we hope that this analysis is useful for folks interested in predicting protein-peptide or domain-SLiM interactions (like us!). and we’re big fans of making rigorous test sets, which gets harder and harder as modeling tasks get more complex.

www.biorxiv.org/content/10.1...

www.biorxiv.org

lindsey guan @lindseyguan.bsky.social · Jun 24

and in the few cases of truly novel complexes, success was often explained by pairing common secondary structure elements together. e.g., AF3 prediction of PDB 8HDJ (5/5)

lindsey guan @lindseyguan.bsky.social · Jun 24

while the target MSA was almost always necessary, we were able to rule out inter-chain covariation as important for good performance by shuffling sequence pairing. e.g., AF2 prediction of PDB 3TRS (4/5)

lindsey guan @lindseyguan.bsky.social · Jun 24

we even found that this bias might even mislead predictions to use wrong but common binding sites. e.g., Boltz-1 prediction of PDB 8V8E (3/5)

lindsey guan @lindseyguan.bsky.social · Jun 24

we found some concerning evidence of memorization for both pre-cutoff structures and structures homologous to pre-cutoff structures 🤔 (2/5)

lindsey guan @lindseyguan.bsky.social · Jun 24

preprint is out for the first project of my phd! we investigated how structure prediction models work for protein-peptide complexes, where the peptide often doesn’t have a deep MSA / there’s no good paired MSA 🧵 (1/5)

www.biorxiv.org/content/10.1...

www.biorxiv.org

Reposted by lindsey guan

bioRxiv Bioinfo @biorxiv-bioinfo.bsky.social · Jun 24

How AlphaFold and related models predict protein-peptide complex structures https://www.biorxiv.org/content/10.1101/2025.06.18.660495v1

Reposted by lindsey guan

David Pfau @davidpfau.com · May 28

The war on science in the US is already having an effect on private sector research like AlphaFold. Bears repeating but the private sector builds on top of things created by academic research for the public good. This hurts everyone.

Reposted by lindsey guan

Eli Lipsitz @eli.lipsitz.net · May 5

There’s been a lot of interest in Game Bub, so I’m going to be launching a crowdfunding campaign through @crowdsupply.bsky.social!

Sign up for updates and to hear when the campaign goes live: www.crowdsupply.com/second-bedro...

This will be an assembled product, not just a kit!

An open-source FPGA retro-emulation handheld

www.crowdsupply.com

Reposted by lindsey guan

@LaBonneLaB @labonnelab.bsky.social · Mar 4

Published in the Chicago Sun Times today

Reposted by lindsey guan

Eli Lipsitz @eli.lipsitz.net · Feb 12

I’m excited to announce Game Bub, an open-source FPGA emulation handheld with support for Game Boy and Game Boy Advance cartridges: eli.lipsitz.net/posts/introd...

Game Bub: open-source FPGA retro emulation handheld

I’m excited to announce the project I’ve been working on for the last year and a half: Game Bub, an open-source FPGA based retro emulation handheld, with support for Game Boy, Game Boy Color, and Game...

eli.lipsitz.net

Reposted by lindsey guan

Carolyn Bertozzi @carolynbertozzi.bskyverified.social · Feb 8

Another assault on US competitiveness at a time when biomedicine is roaring with innovation, an own-goal in a high-stakes international tournament. Of course the real losers are American people needing medicines and cures.

www.washingtonpost.com/health/2025/...

NIH cuts billions of dollars in biomedical funding, effective immediately

The move halts a large slice of money for most universities and research institutions virtually overnight, imperiling vital research in everything from cancer to heart disease.

www.washingtonpost.com

Reposted by lindsey guan

Anne Carpenter @drannecarpenter.bsky.social · Feb 6

So one approach is to omit these words from your research. The other is for all of us to double down on including these words in EVERYTHING.

Dr. Rebecca Calisi-Rodríguez @rebeccacalisi.bsky.social · Feb 6

This will be the biggest game of Taboo the U.S. has ever seen.

Reposted by lindsey guan

anthrofuentes.bsky.social @anthrofuentes.bsky.social · Feb 2

If you can, do something, speak up and act up. www.science.org/doi/10.1126/...

Scientists as political advocates

Science, both teaching and doing, is under attack. The recent US presidential election of a person and platform with anti-science bias exemplifies this. The study of climate processes and patterns and...

www.science.org

Reposted by lindsey guan

Simon Mathis @simonmathis.bsky.social · Nov 25

A weekend project from a while back -- this little package (with no dependencies) allows you to interact with pymol remotely.

I use it a lot for my protein design workflows together with @biotite.bsky.social.

Just `pip install pymol-remote`

lindsey guan @lindseyguan.bsky.social · Nov 27

To learn more, read out the publication in PRX Life and use the code (github.com/MadryLab/rla). Thanks to Foster Birnbaum, Saachi Jain, Aleksander Madry, and Amy E. Keating for this work! (my bsky account is simple a vessel 🦋) 3/3

GitHub - MadryLab/rla: Residue Level Alignment

Residue Level Alignment . Contribute to MadryLab/rla development by creating an account on GitHub.

lindsey guan @lindseyguan.bsky.social · Nov 27

RLA has been tested on several benchmark sets, including several design libraries of miniprotein designs for a variety of protein targets. For all but two targets, filtering with RLA results in a higher success rate after subsequent AF2-based filtering. 2/3

lindsey guan @lindseyguan.bsky.social · Nov 27

✨new work from our lab!✨

RLA is a contrastive-learning approach that assesses sequence-structure compatibility by aligning sequence and structure machine learning representations! RLA can successfully filter protein binder designs. 1/3 🧶

journals.aps.org/prxlife/abst...

Jointly Embedding Protein Structures and Sequences through Residue Level Alignment

Residue Level Alignment integrates protein sequence and structure information in a self-supervised model, improving speed and precision in predicting protein binding and structural stability.

journals.aps.org