mlsanityData to 14 Oct 2026

Repurposing Protein Language Models for Latent Flow-Based Fitness Optimization

arXiv 2602.02425 ↗First posted 2 Feb 2026cs.LG q-bio.QMcited by 0

Not on the map. Why?

Abstract (arXiv)

Protein fitness optimization is challenged by a vast combinatorial landscape where high-fitness variants are extremely sparse. Many current methods either underperform or require computationally expensive gradient-based sampling. We present CHASE, a framework that repurposes the evolutionary knowledge of pretrained protein language models by compressing their embeddings into a compact latent space. By training a conditional flow-matching model with classifier-free guidance, we enable the direct generation of high-fitness variants without predictor-based guidance during the ODE sampling steps. CHASE achieves state-of-the-art performance on AAV and GFP protein design benchmarks. Finally, we show that bootstrapping with synthetic data can further enhance performance in data-constrained settings.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 20 of its 40 references that are in the map. First posted on arXiv 2 Feb 2026.

Builds on

20 of 40 references.

References · 20

  1. 2022Classifier-Free Diffusion Guidancecited by 5,847
  2. 2022Flow Matching for Generative Modelingcited by 5,702
  3. 2022Diffusion Posterior Sampling for General Noisy Inverse Problemscited by 1,416
  4. 2022Biological Sequence Design with GFlowNetscited by 174
  5. 2023Protein Design with Guided Discrete Diffusioncited by 157
  6. 2023The geometry of hidden representations of large transformer modelscited by 112
  7. 2019Conditioning by adaptive sampling for robust designcited by 111
  8. 2021Oops I Took A Gradient: Scalable Sampling for Discrete Distributionscited by 92
  9. 2021Conservative Objective Models for Effective Offline Model-Based Optimizationcited by 88
  10. 2020AdaLead: A simple and robust adaptive greedy search algorithm for sequence designcited by 65
  11. 2017The reparameterization trick for acquisition functionscited by 63
  12. 2023Protein Discovery with Discrete Walk-Jump Samplingcited by 38
8 more
  1. 2025Energy Matching: Unifying Flow Matching and Energy-Based Models for Generative Modelingcited by 27
  2. 2023Improving Protein Optimization with Smoothed Fitness Landscapescited by 21
  3. 2022Plug & Play Directed Evolution of Proteins with Gradient-based Discrete MCMCcited by 17
  4. 2024Robust Optimization in Protein Fitness Landscapes Using Reinforcement Learning in Latent Spacecited by 13
  5. 2020A primer on model-guided exploration of fitness landscapes for biological sequence designcited by 7
  6. 2025Temporal Score Rescaling for Temperature Sampling in Diffusion and Flow Modelscited by 7
  7. 2025A Variational Perspective on Generative Protein Fitness Optimizationcited by 4
  8. 2025ProtFlow: Fast Protein Sequence Design via Flow Matching on Compressed Protein Language Model Embeddingscited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Steering Generative Models with Experimental Data for Protein Fitness Optimization
  2. 2025Evolutionary Profiles for Protein Fitness Prediction
  3. 2026EvoFlows: Evolutionary Edit-Based Flow-Matching for Protein Engineering
  4. 2026No Generation without Representation: Efficient Causal Protein Language Models Enable Zero-Shot Fitness Estimation
  5. 2025Sparse Autoencoders for Low-N Protein Function Prediction and Design
  6. 2025From Supervision to Exploration: What Does Protein Language Model Learn During Reinforcement Learning?
  7. 2026Be Your Own Teacher: Steering Protein Language Models via Unsupervised Reward Optimization
  8. 2025Boosting In-Silicon Directed Evolution with Fine-Tuned Protein Language Model and Tree Search