Nathaniel Blalock

Let me know if you’d like me to clarify anything. I’m happy to talk!

Me too 🤪 It is really exciting to be submitting! We definitely learned a lot along the way

Reinforcement learning with experimental feedback (RLXF) shifts protein language models so that they generate sequences with improved properties

@nathanielblalock.bsky.social @philromero.bsky.social

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

Thank you for sharing our work @kevinkaichuang.bsky.social! It means a lot

Thank you for posting about our preprint!

GitHub - RomeroLab/RLXF: Consolidated repository to perform RLXF Consolidated repository to perform RLXF. Contribute to RomeroLab/RLXF development by creating an account on GitHub.

and our open-source code at github.com/RomeroLab/RLXF

Functional alignment of protein language models via reinforcement learning Protein language models (pLMs) enable generative design of novel protein sequences but remain fundamentally misaligned with protein engineering goals, as they lack explicit understanding of function a...

Want to learn more? Check out our preprint at www.biorxiv.org/content/10.1...

We apply RLXF across five diverse protein classes to demonstrate its generalizability and effectiveness at generating optimized sequences by learning functional constraints beyond those captured during pre-training

Experimental validation reveals the RLXF-aligned model generates a higher fraction of functional sequences, a greater number of sequences more fluorescent than CreiLOV, and the brightest oxygen-independent fluorescent protein variant reported to date

We align ESM-2 to experimental fluorescence data from the CreiLOV flavin-binding fluorescent protein. The aligned model learns to prioritize mutations that enhance fluorescence, many of which are missed by the base model

RLXF follows a two-phase strategy inspired by RLHF. Supervised Fine-Tuning initializes the model in the right region of sequence space. Proximal Policy Optimization directly aligns sequence generation with feedback from a reward function like a sequence-function predictor

Pre-trained pLMs generate highly diverse sequences mirroring statistical patterns from natural proteins. But here's the challenge: they lack an explicit understanding of function, often failing to generate proteins with enhanced or non-natural activities. RLXF bridges this gap!

We are excited in the @philromero.bsky.social lab to share our new preprint introducing RLXF for the functional alignment of protein language models (pLMs) with experimentally derived notions of biomolecular function!

Why we do research - College of Engineering - University of Wisconsin-Madison At a time when the role and value of higher education are being questioned, it's imperative to reflect on the important benefits of research.

Great article, simple reminder about the value of higher education! engineering.wisc.edu/blog/why-we-...

Scalable and cost-efficient custom gene library assembly from oligopools Advances in metagenomics, deep learning, and generative protein design have enabled broad in silico exploration of sequence space, but experimental characterization is still constrained by the cost an...

🎉Congrats to Chase on her new preprint! She developed OMEGA--a simple method for assembling custom gene panels for as little as $1.50 per gene. Big step forward protein engineering and design!🧬
www.biorxiv.org/content/10.1...

Post the amazing science things you have done with federal funding.

It was a pleasure meeting you! Y'all are doing super interesting and relevant work. It will be cool to see how we can continue to interact and maybe collaborate in the future!

Favorite foods! Tandoori chicken and chili momo's: everestkitchen.ca. Onigiri! www.onigiriya.ca. Pho: www.viethouserestaurant.com.

From Distributional to Overton Pluralism: Investigating Large Language Model Alignment The alignment process changes several properties of a large language model's (LLM's) output distribution. We analyze two aspects of post-alignment distributional shift of LLM responses. First, we re-e...

Papers #4: arxiv.org/abs/2406.17692 from the incredible
@gregdnlp.bsky.social. I really like how explore what happens during the alignment of LLM's with RLHF. This was so cool to see having observed similar outcomes in my research.

Self-Play Fine-Tuning of Diffusion Models for Text-to-Image Generation Fine-tuning Diffusion Models remains an underexplored frontier in generative artificial intelligence (GenAI), especially when compared with the remarkable progress made in fine-tuning Large Language M...

Papers #2-3: arxiv.org/abs/2402.10210 and arxiv.org/abs/2405.00675 from the incredible
@quanquangu.bsky.social. I really like how they explore new techniques for RLHF

Guiding Generative Protein Language Models with Reinforcement Learning Autoregressive protein language models (pLMs) have emerged as powerful tools to efficiently design functional proteins with extraordinary diversity, as evidenced by the successful generation of divers...

Paper #1: arxiv.org/abs/2412.12979
Aligning autoregressive pLM's to generate EGFR binders via Direct Policy Optimization (DPO) from the incredible @noeliaferruz.bsky.social who gave a great talk as part of the MLSB workshop

My 1st NeurIPS was a wonderful experience - incredible to see so much research in protein design and reinforcement learning. Here are my favorite papers (and favorite places I got food in Vancouver 😋):

Hey Kevin, could I be added? This is really helpful for joining Bluesky! Thank you for doing it

Three BioML starter packs now!

Pack 1: go.bsky.app/2VWBcCd
Pack 2: go.bsky.app/Bw84Hmc
Pack 3: go.bsky.app/NAKYUok

DM if you want to be included (or nominate people who should be!)