Samuel Sledzieski

From systems operators to systems architects Going up a level from data generation to think about the data systems we design and embed

Announcing the Diffuse Project! We're unlocking protein dynamics through diffuse X-ray scattering - the overlooked signal that could revolutionize how we understand protein motion. seemay.substack.com/p/from-syste...

The results of the ISCB Leadership vote are in! 

Terms for the ISCB President-Elect and ISCB Vice President will start on January 21, 2026.

I just got the notice that all the FlyBase people at Harvard, including me, will be laid off on October 12. I'm devastated.

The open-source Masala software suite: Facilitating rapid methods development for synthetic heteropolymer design Although canonical protein design has benefited from machine learning methods trained on databases of protein sequences and structures, synthetic heteropolymer design still relies heavily on physics-b...

We have a new preprint out on the Masala software suite, a free and open-source platform for easy biomolecular modelling methods development. www.biorxiv.org/content/10.1...

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 ...

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🧵

DirectContacts2: A network of direct physical protein interactions derived from high-throughput mass spectrometry experiments Cellular function is driven by the activity proteins in stable complexes. Protein complex assembly depends on the direct physical association of component proteins. Advances in macromolecular structur...

We are excited to share our preprint describing DirectContacts2! Here we develop a machine learning model to discriminate between direct and indirect protein interactions. We use our model to construct a highly accurate wiring diagram of the human cell.

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

Stay tuned for details on the 6th edition of MLSB, officially happening this December in downtown San Diego, CA!

That's a wrap! The results of the first #cryoEM heterogeneity challenge are up on biorxiv!
biorxiv.org/content/10.110

Everyone working in a STEM field should read this - ‘writing is thinking’
www.nature.com/articles/s44...

New Study Reveals Subclasses of Autism by Linking Traits to Genetics New Study Reveals Subclasses of Autism by Linking Traits to Genetics on Simons Foundation

#FlatironCCB scientists Natalie Sauerwald, Olga Troyanskaya and colleagues leveraged @sparkforautism.bsky.social data to reveal four distinct groups that link autism-related traits with underlying genetics. Read more: www.simonsfoundation.org/2025/07/09/n... #science #biology #autismresearch

We hope this update will significantly expand the types of researchers + compute systems that can use D-SCRIPT! Manuscript with benchmarks + implementation details coming hopefully soon, and thanks again so much to Daniel for his hard work pushing this out!

Along with a bunch of other quality-of-life updates (including a mode specifically designed for host x pathogen/symbiont interactions! 🦠), v0.3.0 will power the sorts of comparative genomics and network analyses that we imagined when we first designed D-SCRIPT.

BMPI makes D-SCRIPT fully parallel in both embedding loading and inference, and can distribute inference to an arbitrary number of available GPUs. At the same time, the blocked inference procedure means that memory usage can be decreased arbitrarily low (with a small trade-off in speed).

In v0.3.0, we solve *both* of these issues, enabling efficient inference on both personal computers and multi-GPU HPC systems. The secret? Our new Blocked Multi-GPU Parallel Inference (BMPI) procedure, led by Daniel Schaffer (github.com/schafferde).

This ends up being extremely wasteful, since D-SCRIPT still made predictions in serial on a single GPU; inference was not possible on personal systems, but much of the compute capacity of these large machines was sitting unused. Whole-genome all-by-all prediction could still take upwards of a week!

Despite this, one of the most common user issues we heard about was that pre-loading embeddings achieved speed by trading off massive memory usage, restricting D-SCRIPT inference at whole-genome scale only to large HPC-like machines.

We designed D-SCRIPT to be extremely high-throughput, and the most common use case we've seen is whole-proteome all-by-all prediction (this use case is at the core of our recent 🎼philharmonic work!)

Today we released a major new update to D-SCRIPT! 🚨

TL;DR, whole-proteome protein-protein interaction prediction is now significantly faster *and* less memory intensive 🚅 🧠

You can get started with the new version with `pip install dscript==0.3.0`
💻 github.com/samsledje/D-...
🧵⬇️

The short film "Molecular Duet" brings the dynamic world of microtubules to life. A collaboration between @flatironinstitute.org researcher Mahsa Mofidi and filmmaker Anne Sofie Nørskov. Watch in full: www.simonsfoundation.org/2025/07/16/w... #science #biology #film #FlatironCCB

From #FlatironCCB scientist Bargeen Turzo and filmmaker Grace Zhang, "How to Connect Two Bodies" is an experimental documentary exploring the invisible choreography in meaningful connections of objects across various scales. Watch in full: www.simonsfoundation.org/2025/07/09/w... #science #film

We're excited to release 𝐦𝐑𝐍𝐀𝐁𝐞𝐧𝐜𝐡, a new benchmark suite for mRNA biology containing 10 diverse datasets with 59 prediction tasks, evaluating 18 foundation model families.

Paper: biorxiv.org/content/10.1...
GitHub: github.com/morrislab/mR...
Blog: blank.bio/post/mrnabench

Volume 41 Issue Supplement_1 | Bioinformatics | Oxford Academic Publishes scientific papers and review articles on new developments in bioinformatics and computational biology. Shorter papers report biologically interesting discoveries using computational methods ...

🖥️🧬: The ISMB 2025 proceedings are out! I’m looking forward to seeing many presentations on this work next week :). academic.oup.com/bioinformati...

Today in the journal Science: BioEmu from Microsoft Research AI for Science. This generative deep learning method emulates protein equilibrium ensembles – key for understanding protein function at scale. www.science.org/doi/10.1126/...

Wow, congratulations!🎉

Exclusive: Famed protein structure competition nears end as NIH grant money runs out Agency silent on funding renewal for contest that inspired creation of AIs that predicted how proteins would fold

Exclusive: An international scientific competition widely credited with spurring the development of artificial intelligence for biology appears to be on its deathbed. scim.ag/44ukS90

CASP is the main reason the protein structure prediction technology and research field advanced over the last 30 years. And the main reason AI based methods have been accepted and widely applied in biology. So shortsighted of NIH to postpone or even halt funding. John Moult is a scientific hero.

Watch: 6 Science Short Films Created by Flatiron Institute Researchers and Filmmakers The two-week-long Symbiosis program yielded six experimental short films about basic science.

Watch: 6 Science Short Films Created by Flatiron Institute Researchers and Filmmakers www.simonsfoundation.org/2025/06/18/w...

One thing that really bothers me with the new "virtual cell" terminology is that it is currently largely focused on a very narrow definition of models that can predict effects of trans perturbations (gene dosage, drugs etc) on gene expression. 1/

Extremely cool work here!

You can start using RocketSHP 🚀 now on GitHub, and we're actively working on better models, improved analysis tools, and ways to link prediction/regression-type models with sampling-based approaches. And if you have any large-scale complex trajectories you want to share, my DMs are always open! 😁